feat(providers): add GPT-5.6 model family (#6862)

* feat(providers): add GPT-5.6 model family

* fix(chatgpt-web): resume temporary chat handoffs

* fix(codex): auto-merge discovery, filter denylist, revalidate on lifecycle

Restore live/GitHub auto-merge for Codex catalogs, drop models via
explicit denylist (GPT-5.4 family), and run scrub+live re-sync once on
first-start, app upgrade, or setup completion. Success log:
kill deprecated models complete.

* fix(codex): preserve live catalog reconciliation

Expose remote-only Codex models without dropping user custom entries, and complete lifecycle revalidation only after a successful internal sync. Keep credentialed self-fetches pinned to the active dashboard listener.

---------

Co-authored-by: backryun <backryun@daonlab.local>

.env.example

@@ -981,7 +981,7 @@ GITHUB_OAUTH_CLIENT_ID=Iv1.b507a08c87ecfe98
 # Used by: open-sse/executors/base.ts — buildHeaders() dynamic lookup.
 # Update these when providers release new CLI versions to avoid blocks.
 
-CLAUDE_USER_AGENT="claude-cli/2.1.195 (external, cli)"
+CLAUDE_USER_AGENT="claude-cli/2.1.207 (external, cli)"
 
 # Disable the deterministic tool-name cloak applied on both Anthropic-bound paths
 # (executors/base.ts native OAuth + executors/cliproxyapi.ts CLIProxyAPI) —
@@ -990,7 +990,7 @@ CLAUDE_USER_AGENT="claude-cli/2.1.195 (external, cli)"
 # stream with a misleading 400 out-of-extra-usage placeholder. Set to true to
 # forward the original names verbatim (debugging only).
 # CLAUDE_DISABLE_TOOL_NAME_CLOAK=false
-CODEX_USER_AGENT="codex-cli/0.142.0 (Windows 10.0.26200; x64)"
+CODEX_USER_AGENT="codex-cli/0.144.1 (Windows 10.0.26200; x64)"
 GITHUB_USER_AGENT="GitHubCopilotChat/0.54.0"
 ANTIGRAVITY_USER_AGENT="antigravity/2.0.1 linux/arm64 google-api-nodejs-client/10.3.0"
 KIRO_USER_AGENT="AWS-SDK-JS/3.0.0 kiro-ide/1.0.0"
@@ -1011,7 +1011,7 @@ CURSOR_USER_AGENT="Cursor/3.4"
 
 # Override Codex client version sent in headers independently of the
 # CODEX_USER_AGENT string. Used by: open-sse/config/codexClient.ts.
-# CODEX_CLIENT_VERSION=0.142.0
+# CODEX_CLIENT_VERSION=0.144.1
 
 # Kill-switch to strip non-standard `codex.*` SSE events (e.g. codex.rate_limits)
 # from the Codex Responses stream. These frames break the OpenAI SDK's
@@ -1898,7 +1898,7 @@ APP_LOG_TO_FILE=true
 # OMNIROUTE_TRANSLATION_API_URL=
 # Bearer token for the translation backend (NEVER commit a real key here).
 # OMNIROUTE_TRANSLATION_API_KEY=
-# Model id, e.g. gpt-4o-mini or cx/gpt-5.4-mini.
+# Model id, e.g. gpt-4o-mini or cx/gpt-5.6-sol.
 # OMNIROUTE_TRANSLATION_MODEL=gpt-4o-mini
 # Per-request timeout in milliseconds (default 60000).
 # OMNIROUTE_TRANSLATION_TIMEOUT_MS=60000

changelog.d/features/6862-gpt-5-6-providers.md

@@ -0,0 +1,2 @@
+- **feat(providers):** Add GPT-5.6 support across OpenAI API, Codex, and ChatGPT Web, including Codex Max/Ultra efforts, VS Code metadata, Fast-tier credit accounting, curated live discovery, the Codex 0.144.1 client identity, and correct chat routing for models that also support image generation ([#6862](https://github.com/diegosouzapw/OmniRoute/pull/6862)) - thanks @backryun
+- **chore(providers):** Align emitted Claude Code identity headers, bridge fingerprints, provider profiles, and documented defaults with claude-cli 2.1.207 ([#6862](https://github.com/diegosouzapw/OmniRoute/pull/6862)) - thanks @backryun

config/quality/eslint-suppressions.json

@@ -2304,7 +2304,7 @@
   },
   "tests/unit/vscode-token-routes.test.ts": {
     "@typescript-eslint/no-explicit-any": {
-      "count": 76
+      "count": 65
     }
   },
   "tests/unit/web-runtime-env.test.ts": {

docs/i18n/zh-CN/docs/reference/ENVIRONMENT.md

@@ -480,7 +480,7 @@ process.env[`${PROVIDER_ID}_USER_AGENT`]
 
 | 变量 | 默认值 | 何时更新 |
 | --- | --- | --- |
-| `CLAUDE_USER_AGENT` | `claude-cli/2.1.195 (external, cli)` | Anthropic 发布新的 CLI 版本时 |
+| `CLAUDE_USER_AGENT` | `claude-cli/2.1.207 (external, cli)` | Anthropic 发布新的 CLI 版本时 |
 | `CLAUDE_DISABLE_TOOL_NAME_CLOAK` | `false` | `executors/base.ts` + `executors/cliproxyapi.ts` | 设为 `1`/`true` 可将第三方测试工具的工具名称原封不动地转发到 Anthropic 的两条绑定路径上（原生 OAuth 和 CLIProxyAPI）。默认情况下 executor 会将非 Claude Code 的工具名称确定性别名化（Claude Code 存在规范映射的用规范映射，否则用 PascalCase），并通过 `_toolNameMap` 在响应中还原，从而确保带 snake_case 工具的测试工具不会被视为指纹化第三方客户端而被拒绝。仅供调试。 |
 | `CODEX_USER_AGENT` | `codex-cli/0.142.0 (Windows 10.0.26200; x64)` | OpenAI 更新 Codex CLI 时 |
 | `CODEX_CLIENT_VERSION` | `0.131.0` | 独立于完整 UA 字符串覆盖 Codex 客户端版本 |
@@ -1112,4 +1112,4 @@ CLI_COMPAT_ALL=1
 
 | 变量 | 默认值 | 源文件 | 说明 |
 | --- | --- | --- | --- |
-| `OMNIROUTE_EVAL_CREDENTIALS` | `{}`（空） | `scripts/compression-eval/index.ts` | 运维人员提供的 JSON 凭证，供离线压缩评估 CLI 使用的服务商使用（通过 `JSON.parse` 解析）。未设置时进行试运行。 |
+| `OMNIROUTE_EVAL_CREDENTIALS` | `{}`（空） | `scripts/compression-eval/index.ts` | 运维人员提供的 JSON 凭证，供离线压缩评估 CLI 使用的服务商使用（通过 `JSON.parse` 解析）。未设置时进行试运行。 |

docs/providers/AGENTROUTER.md

@@ -130,7 +130,7 @@ request (see `open-sse/services/claudeCodeCompatible.ts`):
 | Header                                      | Value                                                                                                   |
 | ------------------------------------------- | ------------------------------------------------------------------------------------------------------- |
 | `Authorization`                             | `Bearer <api-key>`                                                                                      |
-| `User-Agent`                                | `claude-cli/2.1.195 (external, sdk-cli)`                                                                |
+| `User-Agent`                                | `claude-cli/2.1.207 (external, sdk-cli)`                                                                |
 | `anthropic-version`                         | `2023-06-01`                                                                                            |
 | `anthropic-beta`                            | `claude-code-20250219,interleaved-thinking-2025-05-14,effort-2025-11-24`                                |
 | Per-connection redact-thinking beta toggle  | Adds `redact-thinking-2026-02-12` for upstreams that specifically require redacted thinking streams     |

docs/reference/ENVIRONMENT.md

@@ -512,7 +512,7 @@ process.env[`${PROVIDER_ID}_USER_AGENT`]
 
 | Variable                         | Default Value                                 | When to Update                                                                                   |
 | -------------------------------- | --------------------------------------------- | ------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| `CLAUDE_USER_AGENT`              | `claude-cli/2.1.195 (external, cli)`          | When Anthropic releases a new CLI version                                                        |
+| `CLAUDE_USER_AGENT`              | `claude-cli/2.1.207 (external, cli)`          | When Anthropic releases a new CLI version                                                        |
 | `CLAUDE_DISABLE_TOOL_NAME_CLOAK` | `false`                                       | `executors/base.ts` + `executors/cliproxyapi.ts`                                                 | Set to `1`/`true` to forward third-party harness tool names verbatim to Anthropic on both Anthropic-bound paths (native OAuth and CLIProxyAPI). By default the executor deterministically aliases non-Claude-Code tool names (Claude Code canonical mapping where one exists, otherwise PascalCase) and reverses them on the response via `_toolNameMap`, so harnesses with snake_case tools are not refused as fingerprinted third-party clients. Debugging only. |
 | `CODEX_USER_AGENT`               | `codex-cli/0.142.0 (Windows 10.0.26200; x64)` | When OpenAI updates the Codex CLI                                                                |
 | `CODEX_CLIENT_VERSION`           | `0.131.0`                                     | Override Codex client version independently of full UA string                                    |

docs/security/STEALTH_GUIDE.md

@@ -88,7 +88,7 @@ Applied to: `system` blocks, all `messages[].content`, and `tools[].description`
 
 For third-party Anthropic relays that only accept "real Claude Code" traffic:
 
-- `CLAUDE_CODE_COMPATIBLE_USER_AGENT = "claude-cli/2.1.195 (external, sdk-cli)"`
+- `CLAUDE_CODE_COMPATIBLE_USER_AGENT = "claude-cli/2.1.207 (external, sdk-cli)"`
 - `CLAUDE_CODE_COMPATIBLE_STAINLESS_PACKAGE_VERSION = "0.94.0"`
 - `CLAUDE_CODE_COMPATIBLE_STAINLESS_RUNTIME_VERSION = "v24.3.0"`
 - `anthropic-beta = "claude-code-20250219,interleaved-thinking-2025-05-14,effort-2025-11-24"` by default
@@ -214,7 +214,7 @@ All MITM endpoints require management auth (`requireCliToolsAuth`). The sudo pas
 
 | Variable                 | Default                                                         |
 | ------------------------ | --------------------------------------------------------------- |
-| `CLAUDE_USER_AGENT`      | `claude-cli/2.1.195 (external, cli)`                            |
+| `CLAUDE_USER_AGENT`      | `claude-cli/2.1.207 (external, cli)`                            |
 | `CODEX_USER_AGENT`       | `codex-cli/0.142.0 (Windows 10.0.26200; x64)`                   |
 | `GITHUB_USER_AGENT`      | `GitHubCopilotChat/0.54.0`                                      |
 | `ANTIGRAVITY_USER_AGENT` | `antigravity/2.0.1 linux/arm64 google-api-nodejs-client/10.3.0` |

open-sse/config/anthropicHeaders.ts

@@ -121,7 +121,7 @@ export function normalizeAnthropicHeaderVariants(headers: Record<string, string>
   }
 }
 
-export const CLAUDE_CLI_VERSION = "2.1.195";
+export const CLAUDE_CLI_VERSION = "2.1.207";
 export const CLAUDE_CLI_USER_AGENT = `claude-cli/${CLAUDE_CLI_VERSION} (external, cli)`;
 export const CLAUDE_CLI_STAINLESS_PACKAGE_VERSION = "0.94.0";
 export const CLAUDE_CLI_STAINLESS_RUNTIME_VERSION = "v24.3.0";

open-sse/config/codexClient.ts

@@ -1,4 +1,4 @@
-const DEFAULT_CODEX_CLIENT_VERSION = "0.144.0";
+const DEFAULT_CODEX_CLIENT_VERSION = "0.144.1";
 const DEFAULT_CODEX_USER_AGENT_PLATFORM = "Windows 10.0.26200";
 const DEFAULT_CODEX_USER_AGENT_ARCH = "x64";
 const CODEX_VERSION_OVERRIDE_ENV = "CODEX_CLIENT_VERSION";

open-sse/config/glmProvider.ts

@@ -150,7 +150,7 @@ export const GLMT_REQUEST_DEFAULTS = Object.freeze({
 });
 
 export const GLM_COUNT_TOKENS_TIMEOUT_MS = 3_000;
-export const GLM_CLAUDE_CODE_USER_AGENT = "claude-cli/2.1.195 (external, sdk-cli)";
+export const GLM_CLAUDE_CODE_USER_AGENT = "claude-cli/2.1.207 (external, sdk-cli)";
 export const GLM_ANTHROPIC_BETA = [
   "claude-code-20250219",
   "interleaved-thinking-2025-05-14",

open-sse/config/imageRegistry.ts

@@ -149,7 +149,11 @@ export const IMAGE_PROVIDERS: Record<string, ImageProviderConfig> = {
     authType: "oauth",
     authHeader: "bearer",
     format: "codex-responses",
-    models: [{ id: "gpt-5.5", name: "GPT 5.5 (Codex Image)" }],
+    models: [
+      { id: "gpt-5.6-sol", name: "GPT 5.6 Sol (Codex Image)" },
+      { id: "gpt-5.6-terra", name: "GPT 5.6 Terra (Codex Image)" },
+      { id: "gpt-5.6-luna", name: "GPT 5.6 Luna (Codex Image)" },
+    ],
     supportedSizes: ["1024x1024", "1024x1536", "1536x1024"],
   },
 
@@ -160,7 +164,7 @@ export const IMAGE_PROVIDERS: Record<string, ImageProviderConfig> = {
     authType: "apikey",
     authHeader: "cookie",
     format: "chatgpt-web",
-    models: [{ id: "gpt-5.3-instant", name: "GPT-5.3 Instant (ChatGPT Web Image)" }],
+    models: [{ id: "gpt-5.5", name: "GPT-5.5 Instant (ChatGPT Web Image)" }],
     supportedSizes: ["1024x1024", "1024x1536", "1536x1024"],
   },
 

open-sse/config/providerRegistry.ts

@@ -17,7 +17,6 @@ import {
   ALIBABA_DASHSCOPE_MODELS,
   GPT_5_5_CONTEXT_LENGTH,
   GPT_5_5_CODEX_CAPABILITIES,
-  GPT_5_4_CODEX_CAPABILITIES,
   CHAT_OPENAI_COMPAT_MODELS,
   mapStainlessOs,
   mapStainlessArch,

open-sse/config/providers/registry/chatgpt-web/index.ts

@@ -9,15 +9,12 @@ export const chatgpt_webProvider: RegistryEntry = {
   authType: "apikey",
   authHeader: "cookie",
   models: [
-    { id: "gpt-5.5-pro", name: "GPT-5.5 Pro" }, // pro tier only, standard effort
+    { id: "gpt-5.6-pro", name: "GPT-5.6 Pro" }, // pro tier only, standard effort
+    { id: "gpt-5.6-thinking", name: "GPT-5.6 Thinking" }, // plus, pro tier
     { id: "gpt-5.5-pro-extended", name: "GPT-5.5 Pro Extended" }, // pro tier only, extended effort
+    { id: "gpt-5.5-pro", name: "GPT-5.5 Pro" }, // pro tier only, standard effort
     { id: "gpt-5.5-thinking", name: "GPT-5.5 Thinking" }, // plus, pro tier
     { id: "gpt-5.5", name: "GPT-5.5 Instant" }, // free, plus, pro tier
-    { id: "gpt-5.4-pro", name: "GPT-5.4 Pro" }, // pro tier only
-    { id: "gpt-5.4-thinking", name: "GPT-5.4 Thinking" }, // plus, pro tier
-    { id: "gpt-5.4-thinking-mini", name: "GPT-5.4 Thinking Mini" }, // free-login only
-    { id: "gpt-5.3", name: "GPT-5.3 Instant" }, // free, free-login, plus, pro tier
-    { id: "gpt-5.3-mini", name: "GPT-5.3 Mini" }, // limit fallback
     { id: "o3", name: "o3" }, // plus ~ tier
   ],
 };

open-sse/config/providers/registry/codex/index.ts

@@ -1,7 +1,7 @@
 import type { RegistryEntry } from "../../shared.ts";
 import {
+  GPT_5_6_CODEX_CAPABILITIES,
   GPT_5_5_CODEX_CAPABILITIES,
-  GPT_5_4_CODEX_CAPABILITIES,
   getCodexDefaultHeaders,
   resolvePublicCred,
 } from "../../shared.ts";
@@ -24,6 +24,106 @@ export const codexProvider: RegistryEntry = {
     tokenUrl: "https://auth.openai.com/oauth/token",
   },
   models: [
+    {
+      id: "gpt-5.6-sol",
+      name: "GPT 5.6 Sol",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-sol-ultra",
+      name: "GPT 5.6 Sol (Ultra)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-sol-max",
+      name: "GPT 5.6 Sol (Max)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-sol-xhigh",
+      name: "GPT 5.6 Sol (xHigh)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-sol-high",
+      name: "GPT 5.6 Sol (High)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-sol-medium",
+      name: "GPT 5.6 Sol (Medium)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-sol-low",
+      name: "GPT 5.6 Sol (Low)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-terra",
+      name: "GPT 5.6 Terra",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-terra-ultra",
+      name: "GPT 5.6 Terra (Ultra)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-terra-max",
+      name: "GPT 5.6 Terra (Max)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-terra-xhigh",
+      name: "GPT 5.6 Terra (xHigh)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-terra-high",
+      name: "GPT 5.6 Terra (High)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-terra-medium",
+      name: "GPT 5.6 Terra (Medium)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-terra-low",
+      name: "GPT 5.6 Terra (Low)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-luna",
+      name: "GPT 5.6 Luna",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-luna-max",
+      name: "GPT 5.6 Luna (Max)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-luna-xhigh",
+      name: "GPT 5.6 Luna (xHigh)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-luna-high",
+      name: "GPT 5.6 Luna (High)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-luna-medium",
+      name: "GPT 5.6 Luna (Medium)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
+    {
+      id: "gpt-5.6-luna-low",
+      name: "GPT 5.6 Luna (Low)",
+      ...GPT_5_6_CODEX_CAPABILITIES,
+    },
     // gpt-5.5 codex OAuth backend caps context at 400K (not the public-API
     // 1.05M). Public refs : openai/codex#19208, #19319, #19464 ;
     // opencode#24171. max_output_tokens is stripped server-side
@@ -77,39 +177,6 @@ export const codexProvider: RegistryEntry = {
       maxInputTokens: 272000,
       maxOutputTokens: 128000,
     },
-    {
-      id: "gpt-5.4",
-      name: "GPT 5.4",
-      ...GPT_5_4_CODEX_CAPABILITIES,
-    },
-    {
-      id: "gpt-5.4-xhigh",
-      name: "GPT 5.4 (xHigh)",
-      ...GPT_5_4_CODEX_CAPABILITIES,
-    },
-    {
-      id: "gpt-5.4-high",
-      name: "GPT 5.4 (High)",
-      ...GPT_5_4_CODEX_CAPABILITIES,
-    },
-    {
-      id: "gpt-5.4-medium",
-      name: "GPT 5.4 (Medium)",
-      ...GPT_5_4_CODEX_CAPABILITIES,
-    },
-    {
-      id: "gpt-5.4-low",
-      name: "GPT 5.4 (Low)",
-      ...GPT_5_4_CODEX_CAPABILITIES,
-    },
-    { id: "gpt-5.4-mini", name: "GPT 5.4 Mini", targetFormat: "openai-responses" },
     { id: "gpt-5.3-codex-spark", name: "GPT 5.3 Codex Spark" },
-    {
-      id: "gpt-5.3-codex",
-      name: "GPT 5.3 Codex",
-      targetFormat: "openai-responses",
-      supportsReasoning: true,
-      supportsXHighEffort: true,
-    },
   ],
 };

open-sse/config/providers/registry/openai/index.ts

@@ -1,5 +1,5 @@
 import type { RegistryEntry } from "../../shared.ts";
-import { REASONING_UNSUPPORTED } from "../../shared.ts";
+import { GPT_5_6_API_CAPABILITIES, REASONING_UNSUPPORTED } from "../../shared.ts";
 
 export const openaiProvider: RegistryEntry = {
   id: "openai",
@@ -11,6 +11,10 @@ export const openaiProvider: RegistryEntry = {
   authHeader: "bearer",
   defaultContextLength: 128000,
   models: [
+    { id: "gpt-5.6", name: "GPT-5.6", ...GPT_5_6_API_CAPABILITIES },
+    { id: "gpt-5.6-sol", name: "GPT-5.6 Sol", ...GPT_5_6_API_CAPABILITIES },
+    { id: "gpt-5.6-terra", name: "GPT-5.6 Terra", ...GPT_5_6_API_CAPABILITIES },
+    { id: "gpt-5.6-luna", name: "GPT-5.6 Luna", ...GPT_5_6_API_CAPABILITIES },
     { id: "gpt-5.5", name: "GPT-5.5", contextLength: 1050000 },
     // #5842: *-pro reasoning models are responses-only upstream — /v1/chat/completions
     // 404s ("only supported in v1/responses"). targetFormat routes them natively.
@@ -37,7 +41,17 @@ export const openaiProvider: RegistryEntry = {
     { id: "gpt-4o", name: "GPT-4o", contextLength: 128000 },
     { id: "gpt-4o-mini", name: "GPT-4o Mini", contextLength: 128000 },
     { id: "o3", name: "O3", contextLength: 200000, unsupportedParams: REASONING_UNSUPPORTED },
-    { id: "o3-mini", name: "O3 Mini", contextLength: 200000, unsupportedParams: REASONING_UNSUPPORTED },
-    { id: "o4-mini", name: "O4 Mini", contextLength: 200000, unsupportedParams: REASONING_UNSUPPORTED },
+    {
+      id: "o3-mini",
+      name: "O3 Mini",
+      contextLength: 200000,
+      unsupportedParams: REASONING_UNSUPPORTED,
+    },
+    {
+      id: "o4-mini",
+      name: "O4 Mini",
+      contextLength: 200000,
+      unsupportedParams: REASONING_UNSUPPORTED,
+    },
   ],
 };

open-sse/config/providers/shared.ts

@@ -284,13 +284,27 @@ export const GPT_5_5_CODEX_CAPABILITIES = {
   contextLength: GPT_5_5_CONTEXT_LENGTH,
 } as const;
 
-export const GPT_5_4_CODEX_CAPABILITIES = {
+// Public OpenAI API limits. These differ from the Codex OAuth catalog limits below.
+export const GPT_5_6_API_CAPABILITIES = {
+  toolCalling: true,
+  supportsReasoning: true,
+  supportsVision: true,
+  supportsXHighEffort: true,
+  contextLength: 1050000,
+  maxInputTokens: 922000,
+  maxOutputTokens: 128000,
+} as const;
+
+// Codex's live catalog reports a 372K usable input budget for GPT-5.6.
+// Keep the reserved 128K output budget explicit, matching the GPT-5.5 catalog contract.
+export const GPT_5_6_CODEX_CAPABILITIES = {
   targetFormat: "openai-responses",
   toolCalling: true,
   supportsReasoning: true,
   supportsVision: true,
   supportsXHighEffort: true,
-  contextLength: 200000,
+  contextLength: 500000,
+  maxInputTokens: 372000,
   maxOutputTokens: 128000,
 } as const;
 

open-sse/executors/chatgpt-web.ts

@@ -34,6 +34,7 @@ import {
 } from "../services/chatgptImageCache.ts";
 import { isThinkingCapableModel, resolveChatGptModel } from "./chatgpt-web/models.ts";
 import { cleanChatGptText } from "./chatgpt-web/citations.ts";
+import { resumeChatGptHandoff, type FinalAssistantAnswer } from "./chatgpt-web/handoff.ts";
 
 // ─── Constants ──────────────────────────────────────────────────────────────
 
@@ -84,8 +85,7 @@ function deviceIdFor(cookie: string): string {
 // OmniRoute model ID → ChatGPT internal slug. The public ChatGPT Web catalog
 // keeps OmniRoute's historical dot-form IDs (e.g. "gpt-5.5-pro"), while
 // ChatGPT's backend routes use dash-form slugs (e.g. "gpt-5-5-pro"). The slug
-// catalog comes from /backend-api/models on a logged-in account;
-// "gpt-5-4-t-mini" is ChatGPT's abbreviated slug for "GPT-5.4 Thinking Mini".
+// catalog comes from /backend-api/models on a logged-in account.
 
 // ─── Browser-like default headers ──────────────────────────────────────────
 
@@ -1069,6 +1069,7 @@ interface ChatGptStreamEvent {
   conversation_id?: string;
   error?: string | { message?: string; code?: string };
   type?: string;
+  token?: string;
   v?: unknown;
 }
 
@@ -1180,6 +1181,8 @@ interface ContentChunk {
   imageGenAsync?: boolean;
   /** True when ChatGPT handed the turn off to a long-running worker. */
   handoff?: boolean;
+  /** Short-lived conduit token used to resume a Temporary Chat handoff. */
+  resumeToken?: string;
 }
 
 interface ImagePointerRef {
@@ -1238,6 +1241,7 @@ async function* extractContent(
   // WebSocket / polling — caller handles that.
   let imageGenAsync = false;
   let handoff = false;
+  let resumeToken: string | null = null;
 
   for await (const event of readChatGptSseEvents(eventStream, signal)) {
     if (event.error) {
@@ -1251,11 +1255,17 @@ async function* extractContent(
 
     if (event.conversation_id) conversationId = event.conversation_id;
 
+    if (event.type === "resume_conversation_token") {
+      if (typeof event.token === "string" && event.token) resumeToken = event.token;
+      continue;
+    }
+
     if (event.type === "stream_handoff") {
       handoff = true;
       yield {
         conversationId: conversationId ?? undefined,
         handoff: true,
+        resumeToken: resumeToken ?? undefined,
       };
       continue;
     }
@@ -1273,8 +1283,7 @@ async function* extractContent(
     // on a tool-role message (handled below).
     if (event.type === "server_ste_metadata") {
       const meta = (event as Record<string, unknown>).metadata as
-        | Record<string, unknown>
-        | undefined;
+        Record<string, unknown> | undefined;
       if (meta && meta.turn_use_case === "image gen") {
         imageGenAsync = true;
       }
@@ -1373,6 +1382,7 @@ async function* extractContent(
     imagePointers: imagePointers.size > 0 ? Array.from(imagePointers.values()) : undefined,
     imageGenAsync,
     handoff,
+    resumeToken: resumeToken ?? undefined,
     done: true,
   };
 }
@@ -1398,13 +1408,6 @@ interface ChatGptConversationDetail {
   mapping?: Record<string, { message?: ChatGptDetailMessage | null }>;
 }
 
-interface FinalAssistantAnswer {
-  text: string;
-  messageId?: string;
-  metadata?: Record<string, unknown>;
-  finished: boolean;
-}
-
 function textFromContentPart(part: unknown): string {
   if (typeof part === "string") return part;
   if (!part || typeof part !== "object") return "";
@@ -1655,10 +1658,11 @@ function buildStreamingResponse(
   // stream finishes without an image_asset_pointer. The executor passes a
   // closure here that knows how to poll the conversation endpoint.
   pollAsyncImage: ((conversationId: string) => Promise<ImagePointerRef[]>) | null,
-  // Optional poller for GPT-5.5 Pro's stream_handoff path. Inline text keeps
-  // streaming as-is; once ChatGPT hands off, we append the final assistant
-  // answer fetched from the conversation detail endpoint. Text requests stay
-  // in Temporary Chat, so these polls should not create sidebar/history items.
+  // Native Temporary Chat handoff continuation. ChatGPT provides a short-lived
+  // conduit token, which resumes the turn without saving it to chat history.
+  resumeFinalAnswer:
+    ((conversationId: string, resumeToken: string) => Promise<FinalAssistantAnswer | null>) | null,
+  // Legacy fallback for handoffs that omit the conduit token.
   pollFinalAnswer: ((conversationId: string) => Promise<FinalAssistantAnswer | null>) | null,
   log: { warn?: (tag: string, msg: string) => void } | null,
   signal?: AbortSignal | null
@@ -1688,6 +1692,7 @@ function buildStreamingResponse(
           let imagePointers: ImagePointerRef[] | undefined;
           let imageGenAsync = false;
           let handoff = false;
+          let resumeToken: string | null = null;
           let emittedText = "";
           let polledFinalAnswer: FinalAssistantAnswer | null = null;
           let parentCandidateMessageId: string | null = null;
@@ -1774,6 +1779,7 @@ function buildStreamingResponse(
             if (chunk.conversationId) conversationId = chunk.conversationId;
             if (chunk.messageId) parentCandidateMessageId = chunk.messageId;
             if (chunk.handoff) handoff = true;
+            if (chunk.resumeToken) resumeToken = chunk.resumeToken;
             if (chunk.error) {
               controller.enqueue(
                 encoder.encode(
@@ -1801,6 +1807,7 @@ function buildStreamingResponse(
               imagePointers = chunk.imagePointers;
               imageGenAsync = chunk.imageGenAsync ?? false;
               handoff = handoff || (chunk.handoff ?? false);
+              if (chunk.resumeToken) resumeToken = chunk.resumeToken;
               if (chunk.messageId) parentCandidateMessageId = chunk.messageId;
               break;
             }
@@ -1810,7 +1817,20 @@ function buildStreamingResponse(
             }
           }
 
-          if (pollFinalAnswer && conversationId && handoff) {
+          if (resumeFinalAnswer && conversationId && handoff && resumeToken) {
+            const stopHb = startHeartbeat();
+            try {
+              const resumed = await resumeFinalAnswer(conversationId, resumeToken);
+              if (resumed?.text) {
+                polledFinalAnswer = resumed;
+                if (resumed.messageId) parentCandidateMessageId = resumed.messageId;
+              }
+            } finally {
+              stopHb();
+            }
+          }
+
+          if (!polledFinalAnswer && pollFinalAnswer && conversationId && handoff) {
             const stopHb = startHeartbeat();
             try {
               const polled = await pollFinalAnswer(conversationId);
@@ -1991,6 +2011,8 @@ async function buildNonStreamingResponse(
   currentMsg: string,
   resolver: ImageResolver | null,
   pollAsyncImage: ((conversationId: string) => Promise<ImagePointerRef[]>) | null,
+  resumeFinalAnswer:
+    ((conversationId: string, resumeToken: string) => Promise<FinalAssistantAnswer | null>) | null,
   pollFinalAnswer: ((conversationId: string) => Promise<FinalAssistantAnswer | null>) | null,
   log: { warn?: (tag: string, msg: string) => void } | null,
   signal?: AbortSignal | null
@@ -2000,6 +2022,7 @@ async function buildNonStreamingResponse(
   let imagePointers: ImagePointerRef[] | undefined;
   let imageGenAsync = false;
   let handoff = false;
+  let resumeToken: string | null = null;
   let answerMetadata: Record<string, unknown> | undefined;
   let parentCandidateMessageId: string | null = null;
 
@@ -2007,6 +2030,7 @@ async function buildNonStreamingResponse(
     if (chunk.conversationId) conversationId = chunk.conversationId;
     if (chunk.messageId) parentCandidateMessageId = chunk.messageId;
     if (chunk.handoff) handoff = true;
+    if (chunk.resumeToken) resumeToken = chunk.resumeToken;
     if (chunk.error) {
       return new Response(
         JSON.stringify({
@@ -2021,6 +2045,7 @@ async function buildNonStreamingResponse(
       imagePointers = chunk.imagePointers;
       imageGenAsync = chunk.imageGenAsync ?? false;
       handoff = handoff || (chunk.handoff ?? false);
+      if (chunk.resumeToken) resumeToken = chunk.resumeToken;
       if (chunk.messageId) parentCandidateMessageId = chunk.messageId;
       break;
     }
@@ -2030,7 +2055,22 @@ async function buildNonStreamingResponse(
     }
   }
 
-  if (pollFinalAnswer && conversationId && (handoff || !fullAnswer.trim())) {
+  let resumedAnswer: FinalAssistantAnswer | null = null;
+  if (resumeFinalAnswer && conversationId && handoff && resumeToken) {
+    resumedAnswer = await resumeFinalAnswer(conversationId, resumeToken);
+    if (resumedAnswer?.text) {
+      fullAnswer = resumedAnswer.text;
+      answerMetadata = resumedAnswer.metadata ?? answerMetadata;
+      if (resumedAnswer.messageId) parentCandidateMessageId = resumedAnswer.messageId;
+    }
+  }
+
+  if (
+    !resumedAnswer?.text &&
+    pollFinalAnswer &&
+    conversationId &&
+    (handoff || !fullAnswer.trim())
+  ) {
     const polled = await pollFinalAnswer(conversationId);
     if (polled?.text) {
       fullAnswer = polled.text;
@@ -2718,8 +2758,7 @@ export class ChatGptWebExecutor extends BaseExecutor {
     clientHeaders,
   }: ExecuteInput) {
     const messages = (body as Record<string, unknown> | null)?.messages as
-      | Array<Record<string, unknown>>
-      | undefined;
+      Array<Record<string, unknown>> | undefined;
     if (!messages || !Array.isArray(messages) || messages.length === 0) {
       return {
         response: errorResponse(400, "Missing or empty messages array"),
@@ -2819,6 +2858,7 @@ export class ChatGptWebExecutor extends BaseExecutor {
     // browser does on page load. Failures here are non-fatal; the worst case
     // is Sentinel still escalates to Turnstile.
     const sessionId = randomUUID();
+    const turnTraceId = randomUUID();
     const deviceId = deviceIdFor(cookie);
     await runSessionWarmup(
       tokenEntry.accessToken,
@@ -2963,6 +3003,7 @@ export class ChatGptWebExecutor extends BaseExecutor {
       Accept: "text/event-stream",
       Authorization: `Bearer ${tokenEntry.accessToken}`,
       Cookie: buildSessionCookieHeader(cookie),
+      "x-oai-turn-trace-id": turnTraceId,
     };
     if (tokenEntry.accountId) headers["chatgpt-account-id"] = tokenEntry.accountId;
     if (reqs.token) headers["openai-sentinel-chat-requirements-token"] = reqs.token;
@@ -3055,6 +3096,16 @@ export class ChatGptWebExecutor extends BaseExecutor {
     const imageResolver = makeImageResolver(resolverCtx);
     const pollAsyncImage = (conversationId: string) =>
       pollForAsyncImage(conversationId, resolverCtx);
+    const resumeFinalAnswer = (conversationId: string, resumeToken: string) =>
+      resumeChatGptHandoff({
+        conversationId,
+        resumeToken,
+        headers,
+        timeoutMs: configuredProPollTimeoutMs(),
+        signal,
+        log,
+        readContent: extractContent,
+      });
     const pollFinalAnswer = resolvedModel.isPro
       ? (conversationId: string) => pollForFinalAssistantAnswer(conversationId, resolverCtx)
       : null;
@@ -3071,6 +3122,7 @@ export class ChatGptWebExecutor extends BaseExecutor {
         created,
         imageResolver,
         pollAsyncImage,
+        resumeFinalAnswer,
         pollFinalAnswer,
         log,
         signal
@@ -3092,6 +3144,7 @@ export class ChatGptWebExecutor extends BaseExecutor {
         parsed.currentMsg,
         imageResolver,
         pollAsyncImage,
+        resumeFinalAnswer,
         pollFinalAnswer,
         log,
         signal

open-sse/executors/chatgpt-web/handoff.ts

@@ -0,0 +1,154 @@
+import { tlsFetchChatGpt } from "../../services/chatgptTlsClient.ts";
+
+const CONVERSATION_RESUME_URL = "https://chatgpt.com/backend-api/f/conversation/resume";
+const RESUME_OFFSETS = [0, 1, 2] as const;
+
+export interface FinalAssistantAnswer {
+  text: string;
+  messageId?: string;
+  metadata?: Record<string, unknown>;
+  finished: boolean;
+}
+
+interface HandoffContentChunk {
+  answer?: string;
+  messageId?: string;
+  metadata?: Record<string, unknown>;
+  error?: string;
+}
+
+type HandoffContentReader = (
+  eventStream: ReadableStream<Uint8Array>,
+  signal?: AbortSignal | null
+) => AsyncIterable<HandoffContentChunk>;
+
+interface ResumeHandoffOptions {
+  conversationId: string;
+  resumeToken: string;
+  headers: Record<string, string>;
+  timeoutMs: number;
+  signal?: AbortSignal | null;
+  log?: { warn?: (tag: string, message: string) => void } | null;
+  readContent: HandoffContentReader;
+}
+
+interface ResumeAttemptOptions extends Pick<
+  ResumeHandoffOptions,
+  "conversationId" | "timeoutMs" | "signal" | "log" | "readContent"
+> {
+  offset: (typeof RESUME_OFFSETS)[number];
+  resumeHeaders: Record<string, string>;
+}
+
+interface ResumeAttemptResult {
+  answer: FinalAssistantAnswer | null;
+  shouldRetry: boolean;
+}
+
+function stringToStream(text: string): ReadableStream<Uint8Array> {
+  const bytes = new TextEncoder().encode(text);
+  return new ReadableStream<Uint8Array>({
+    start(controller) {
+      controller.enqueue(bytes);
+      controller.close();
+    },
+  });
+}
+
+async function readFinalAssistantAnswer(
+  eventStream: ReadableStream<Uint8Array>,
+  signal: AbortSignal | null | undefined,
+  readContent: HandoffContentReader
+): Promise<FinalAssistantAnswer | null> {
+  let text = "";
+  let messageId: string | undefined;
+  let metadata: Record<string, unknown> | undefined;
+
+  for await (const chunk of readContent(eventStream, signal)) {
+    if (chunk.error) return null;
+    if (chunk.answer) text = chunk.answer;
+    if (chunk.messageId) messageId = chunk.messageId;
+    if (chunk.metadata) metadata = chunk.metadata;
+  }
+
+  if (!text.trim()) return null;
+  return { text, messageId, metadata, finished: true };
+}
+
+async function attemptResumeOffset({
+  conversationId,
+  offset,
+  resumeHeaders,
+  timeoutMs,
+  signal,
+  log,
+  readContent,
+}: ResumeAttemptOptions): Promise<ResumeAttemptResult> {
+  try {
+    const response = await tlsFetchChatGpt(CONVERSATION_RESUME_URL, {
+      method: "POST",
+      headers: resumeHeaders,
+      body: JSON.stringify({ conversation_id: conversationId, offset }),
+      timeoutMs,
+      signal,
+      stream: true,
+    });
+
+    if (response.status === 404) return { answer: null, shouldRetry: true };
+    if (response.status >= 400) {
+      log?.warn?.(
+        "CGPT-WEB",
+        `conversation resume ${response.status}: ${(response.text || "").slice(0, 300)}`
+      );
+      return { answer: null, shouldRetry: false };
+    }
+
+    const eventStream = response.body ?? (response.text ? stringToStream(response.text) : null);
+    if (!eventStream) return { answer: null, shouldRetry: true };
+
+    const answer = await readFinalAssistantAnswer(eventStream, signal, readContent);
+    return { answer, shouldRetry: !answer };
+  } catch (error) {
+    log?.warn?.(
+      "CGPT-WEB",
+      `conversation resume failed: ${error instanceof Error ? error.message : String(error)}`
+    );
+    return { answer: null, shouldRetry: false };
+  }
+}
+
+export async function resumeChatGptHandoff({
+  conversationId,
+  resumeToken,
+  headers,
+  timeoutMs,
+  signal,
+  log,
+  readContent,
+}: ResumeHandoffOptions): Promise<FinalAssistantAnswer | null> {
+  const resumeHeaders = {
+    ...headers,
+    Accept: "text/event-stream",
+    "Content-Type": "application/json",
+    "x-conduit-token": resumeToken,
+    "X-OpenAI-Target-Path": "/backend-api/f/conversation/resume",
+    "X-OpenAI-Target-Route": "/backend-api/f/conversation/resume",
+  };
+
+  for (const offset of RESUME_OFFSETS) {
+    const attempt = await attemptResumeOffset({
+      conversationId,
+      resumeHeaders,
+      offset,
+      timeoutMs,
+      signal,
+      log,
+      readContent,
+    });
+    if (attempt.answer) return attempt.answer;
+    if (!attempt.shouldRetry) return null;
+  }
+
+  log?.warn?.("CGPT-WEB", `conversation resume returned no assistant text for ${conversationId}`);
+  return null;
+}

open-sse/executors/chatgpt-web/models.ts

@@ -3,24 +3,22 @@
 
 export const MODEL_MAP: Record<string, string> = {
   // ChatGPT backend slugs are also accepted directly for power users / tests.
+  "gpt-5-6-pro": "gpt-5-6-pro",
+  "gpt-5-6-thinking": "gpt-5-6-thinking",
   "gpt-5-5-pro": "gpt-5-5-pro",
   "gpt-5-5-pro-extended": "gpt-5-5-pro",
   "gpt-5-5-thinking": "gpt-5-5-thinking",
   "gpt-5-5": "gpt-5-5",
-  "gpt-5-4-pro": "gpt-5-4-pro",
-  "gpt-5-4-thinking": "gpt-5-4-thinking",
-  "gpt-5-4-t-mini": "gpt-5-4-t-mini",
   "gpt-5-3": "gpt-5-3",
   "gpt-5-3-mini": "gpt-5-3-mini",
 
   // Public OmniRoute dot-form ids exposed by the provider catalog.
+  "gpt-5.6-pro": "gpt-5-6-pro",
+  "gpt-5.6-thinking": "gpt-5-6-thinking",
   "gpt-5.5-pro": "gpt-5-5-pro",
   "gpt-5.5-pro-extended": "gpt-5-5-pro",
   "gpt-5.5-thinking": "gpt-5-5-thinking",
   "gpt-5.5": "gpt-5-5",
-  "gpt-5.4-pro": "gpt-5-4-pro",
-  "gpt-5.4-thinking": "gpt-5-4-thinking",
-  "gpt-5.4-thinking-mini": "gpt-5-4-t-mini",
   "gpt-5.3-instant": "gpt-5-3-instant",
   "gpt-5.3": "gpt-5-3",
   "gpt-5.3-mini": "gpt-5-3-mini",
@@ -28,6 +26,8 @@ export const MODEL_MAP: Record<string, string> = {
 };
 
 export const MODEL_FORCED_EFFORT: Record<string, "standard" | "extended"> = {
+  "gpt-5-6-pro": "standard",
+  "gpt-5.6-pro": "standard",
   "gpt-5-5-pro": "standard",
   "gpt-5-5-pro-extended": "extended",
   "gpt-5.5-pro": "standard",
@@ -36,9 +36,7 @@ export const MODEL_FORCED_EFFORT: Record<string, "standard" | "extended"> = {
 
 /** Set of chatgpt.com slugs that the user_last_used_model_config endpoint
  * accepts a `thinking_effort` value for, derived from MODEL_MAP so adding a
- * new thinking entry there automatically extends this set. Includes the
- * abbreviated slug `gpt-5-4-t-mini` (no literal "thinking" substring) — the
- * reason this set exists at all rather than a substring match.
+ * new thinking entry there automatically extends this set.
  *
  * Derived from MODEL_MAP keys (always dot-form) that contain "thinking" or
  * are the `o3` reasoning model; the values are the chatgpt.com-side slugs. */
@@ -52,18 +50,8 @@ export const THINKING_CAPABLE_SLUGS: ReadonlySet<string> = new Set(
  * models and the o-series. PATCHing for a non-thinking surface is a no-op
  * (the server accepts it but the routing-time read picks the wrong knob).
  *
- * Three branches because the input can arrive in three shapes:
- *   1. OmniRoute dot-form id (`gpt-5.4-thinking-mini`) — every thinking
- *      variant carries the literal "thinking" substring here.
- *   2. Resolved chatgpt.com slug containing "thinking" (`gpt-5-5-thinking`).
- *   3. Resolved chatgpt.com slug that drops the substring under abbreviation
- *      (`gpt-5-4-t-mini`). Looked up via THINKING_CAPABLE_SLUGS, which is
- *      derived from MODEL_MAP itself so adding a new abbreviated thinking
- *      mapping automatically extends the check.
- *
- * Branch 3 also catches the case where a caller passes the chatgpt.com slug
- * directly as the `model` field (no MODEL_MAP translation needed), which
- * would otherwise silently bypass the PATCH. */
+ * The lookup also catches callers that pass a chatgpt.com slug directly as
+ * the `model` field without MODEL_MAP translation. */
 export function isThinkingCapableModel(modelId: string, slug: string): boolean {
   return (
     modelId.includes("thinking") ||
@@ -128,6 +116,6 @@ export function resolveChatGptModel(
   const slug = MODEL_MAP[model] ?? model;
   const forcedEffort = MODEL_FORCED_EFFORT[model] ?? null;
   const effort = forcedEffort ?? resolveThinkingEffort(body, providerSpecificData);
-  const isPro = slug === "gpt-5-5-pro";
+  const isPro = slug === "gpt-5-6-pro" || slug === "gpt-5-5-pro";
   return { slug, effort, isPro };
 }

open-sse/executors/claudeIdentity.ts

@@ -12,7 +12,7 @@ import { createHash, randomBytes, randomUUID } from "node:crypto";
 
 // ---------- Versions ------------------------------------------------------
 
-export const CLAUDE_CODE_VERSION = "2.1.195";
+export const CLAUDE_CODE_VERSION = "2.1.207";
 /** Bundled @anthropic-ai/sdk version for the pinned CLI release. */
 export const CLAUDE_CODE_STAINLESS_VERSION = "0.94.0";
 

open-sse/executors/codex.ts

@@ -117,8 +117,11 @@ function codexWebSocketUnavailableResponse(): Response {
 export { getCodexModelScope, getCodexRateLimitKey, type CodexQuotaScope };
 
 // Ordered list of effort levels from lowest to highest
-const EFFORT_ORDER = ["none", "low", "medium", "high", "xhigh"] as const;
+const EFFORT_ORDER = ["none", "low", "medium", "high", "xhigh", "max", "ultra"] as const;
 type EffortLevel = (typeof EFFORT_ORDER)[number];
+const STANDARD_EFFORT_SUFFIXES = ["none", "low", "medium", "high", "xhigh"] as const;
+const GPT_5_6_MAX_ALIAS_MODELS = new Set(["gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna"]);
+const GPT_5_6_ULTRA_ALIAS_MODELS = new Set(["gpt-5.6-sol", "gpt-5.6-terra"]);
 const CODEX_FAST_WIRE_VALUE = "priority";
 const CODEX_RESPONSES_WS_URL = "wss://chatgpt.com/backend-api/codex/responses";
 
@@ -127,7 +130,17 @@ function splitCodexReasoningSuffix(model: unknown): {
   effort: EffortLevel | null;
 } {
   const modelId = typeof model === "string" ? model : "";
-  for (const level of EFFORT_ORDER) {
+  const gpt56AliasMatch = /^(gpt-5\.6-(?:sol|terra|luna))-(max|ultra)$/.exec(modelId);
+  if (gpt56AliasMatch) {
+    const [, baseModel, alias] = gpt56AliasMatch;
+    const supportedModels =
+      alias === "ultra" ? GPT_5_6_ULTRA_ALIAS_MODELS : GPT_5_6_MAX_ALIAS_MODELS;
+    if (supportedModels.has(baseModel)) {
+      return { baseModel, effort: alias as EffortLevel };
+    }
+  }
+
+  for (const level of STANDARD_EFFORT_SUFFIXES) {
     if (modelId.endsWith(`-${level}`)) {
       return {
         baseModel: modelId.slice(0, -`-${level}`.length),
@@ -338,10 +351,13 @@ function normalizeServiceTierValue(value: unknown): string | undefined {
 
 /**
  * Maximum reasoning effort allowed per Codex model.
- * Models not listed here default to "xhigh" (unrestricted).
+ * Models not listed here retain the legacy xhigh cap.
  * Update this table when Codex releases new models with different caps.
  */
 const MAX_EFFORT_BY_MODEL: Record<string, EffortLevel> = {
+  "gpt-5.6-sol": "ultra",
+  "gpt-5.6-terra": "ultra",
+  "gpt-5.6-luna": "max",
   "gpt-5.3-codex": "xhigh",
   "gpt-5.1-codex-max": "xhigh",
   "gpt-5-mini": "high",
@@ -370,7 +386,6 @@ const CODEX_DEFAULT_REASONING_SUMMARY = "auto";
 function normalizeEffortValue(value: unknown): string | undefined {
   if (typeof value !== "string") return undefined;
   const normalized = value.trim().toLowerCase();
-  if (normalized === "max") return "xhigh";
   return normalized || undefined;
 }
 
@@ -587,8 +602,7 @@ function extractCodexSseErrorMessage(text: string, fallback: string): string {
       const parsed = JSON.parse(data) as Record<string, unknown>;
       const directError = parsed.error as Record<string, unknown> | undefined;
       const nestedError = (parsed.response as Record<string, unknown> | undefined)?.error as
-        | Record<string, unknown>
-        | undefined;
+        Record<string, unknown> | undefined;
       const message =
         (typeof directError?.message === "string" && directError.message) ||
         (typeof nestedError?.message === "string" && nestedError.message) ||
@@ -1054,9 +1068,7 @@ export class CodexExecutor extends BaseExecutor {
       headers["chatgpt-account-id"] = workspaceId;
     }
     const clientIdentity = credentials?.providerSpecificData?.codexClientIdentity as
-      | CodexClientIdentity
-      | null
-      | undefined;
+      CodexClientIdentity | null | undefined;
 
     // Originator header — identifies the client type to the Codex backend.
     // Ref: openai/codex login/src/auth/default_client.rs DEFAULT_ORIGINATOR = "codex_cli_rs"
@@ -1295,8 +1307,7 @@ export class CodexExecutor extends BaseExecutor {
       // gpt-5.3-codex-spark (and other Spark-scope models) reject image_generation
       // upstream even on paid-plan accounts, so drop it independent of plan (#6651).
       dropImageGeneration:
-        isCodexFreePlan(credentials?.providerSpecificData) ||
-        getCodexModelScope(model) === "spark",
+        isCodexFreePlan(credentials?.providerSpecificData) || getCodexModelScope(model) === "spark",
       preserveCustomTools: nativeCodexPassthrough,
     });
 
@@ -1335,9 +1346,11 @@ export class CodexExecutor extends BaseExecutor {
       modelEffort || explicitReasoning || requestReasoningEffort || fallbackReasoningEffort;
 
     if (rawEffort) {
+      const clampedEffort = clampEffort(cleanModel, rawEffort);
       body.reasoning = {
         ...(reasoningRecord || {}),
-        effort: clampEffort(cleanModel, rawEffort),
+        // Ultra coordinates delegation in Codex clients; the upstream wire effort is Max.
+        effort: clampedEffort === "ultra" ? "max" : clampedEffort,
       };
     }
     ensureCodexReasoningSummary(body);
@@ -1384,9 +1397,7 @@ export class CodexExecutor extends BaseExecutor {
       applyCodexClientMetadata(
         body,
         credentials?.providerSpecificData?.codexClientIdentity as
-          | CodexClientIdentity
-          | null
-          | undefined
+          CodexClientIdentity | null | undefined
       );
     }
 

open-sse/executors/commandCode.ts

@@ -61,7 +61,7 @@ function normalizeContentText(content: unknown): string {
  * capability per the official CC model registry, but are NOT caught
  * by the shared {@link isVisionModelId} heuristic. Kept as a local
  * set because these are CC-specific model IDs (vendor-prefix shapes
- * like "moonshotai/Kimi-K2.6" or CC aliases like "gpt-5.4-mini").
+ * like "moonshotai/Kimi-K2.6" or CC aliases like "gpt-5.6-luna").
  *
  * Source: Command Code /alpha/generate model registry (docs).
  */
@@ -73,7 +73,7 @@ const CC_VISION_MODEL_PATTERNS: readonly RegExp[] = [
   // Anthropic
   /claude-fable/i, // claude-fable-5 (not covered by claude-opus/sonnet/haiku-4)
   // OpenAI
-  /gpt-5/i, // gpt-5.5, gpt-5.4, gpt-5.3-codex, gpt-5.4-mini
+  /gpt-5/i, // gpt-5.6, gpt-5.5, gpt-5.3-codex
   // Sakana
   /fugu/i, // sakana/fugu-ultra
 ];

open-sse/services/ccBridgeTransforms.ts

@@ -96,7 +96,7 @@ export interface InjectBillingHeaderOp {
    *   - static-zero: emit "00000" (relay endpoints don't validate)
    */
   cchAlgo: "sha256-first-user" | "xxhash64-body" | "static-zero";
-  /** Override the embedded `cc_version=` value. Defaults to `2.1.195`. */
+  /** Override the embedded `cc_version=` value. Defaults to `2.1.207`. */
   version?: string;
 }
 
@@ -114,7 +114,7 @@ export const CCH_SALT = "59cf53e54c78";
 /** Character positions sampled from the first user message text. */
 export const CCH_POSITIONS = [4, 7, 20] as const;
 /** Default `cc_version=` value embedded in the billing header. */
-export const DEFAULT_CLAUDE_CODE_VERSION = "2.1.195";
+export const DEFAULT_CLAUDE_CODE_VERSION = "2.1.207";
 /** Identity sentinel prepended for Claude Agent SDK callers. */
 export const CLAUDE_AGENT_SDK_IDENTITY =
   "You are a Claude agent, built on Anthropic's Claude Agent SDK.";

open-sse/services/claudeCodeCompatible.ts

@@ -42,8 +42,8 @@ export {
   CLAUDE_CODE_COMPATIBLE_REDACT_THINKING_BETA,
   resolveClaudeCodeCompatibleAnthropicBeta,
 } from "./claudeCodeCompatibleBeta.ts";
-export const CLAUDE_CODE_COMPATIBLE_VERSION = "2.1.195";
-export const CLAUDE_CODE_COMPATIBLE_USER_AGENT = "claude-cli/2.1.195 (external, sdk-cli)";
+export const CLAUDE_CODE_COMPATIBLE_VERSION = "2.1.207";
+export const CLAUDE_CODE_COMPATIBLE_USER_AGENT = "claude-cli/2.1.207 (external, sdk-cli)";
 export const CLAUDE_CODE_COMPATIBLE_STAINLESS_PACKAGE_VERSION = "0.94.0";
 export const CLAUDE_CODE_COMPATIBLE_STAINLESS_RUNTIME_VERSION = "v24.3.0";
 export const CONTEXT_1M_BETA_HEADER = "context-1m-2025-08-07";

scripts/check/check-env-doc-sync.mjs

@@ -148,6 +148,9 @@ const IGNORE_FROM_CODE = new Set([
   // write-build-sha.mjs to stamp dist/BUILD_SHA — injected by the build, never
   // configured by users in .env.
   "OMNIROUTE_BUILD_SHA",
+  // Listener-owned self-fetch transport signal. The HTTP/HTTPS launchers set
+  // this before application imports; it is not user-configurable product env.
+  "OMNIROUTE_INTERNAL_SCHEME",
   // Source typo / placeholder.
   "OMNIROUT",
   // Static config alias path (the canonical var is OMNIROUTE_PAYLOAD_RULES_PATH).

scripts/dev/run-next.mjs

@@ -12,10 +12,7 @@ import { ensurePeerStampToken, stampPeerIp } from "./peer-stamp.mjs";
 import methodGuard from "./http-method-guard.cjs";
 import headResponseGuard from "./head-response-guard.cjs";
 import { ensureNativeSqlite } from "./ensure-native-sqlite.mjs";
-import {
-  isTurbopackCacheCorruption,
-  purgeAllTurbopackCaches,
-} from "./turbopackCacheHeal.mjs";
+import { isTurbopackCacheCorruption, purgeAllTurbopackCaches } from "./turbopackCacheHeal.mjs";
 import { randomUUID } from "node:crypto";
 
 const { maybeHandleDisallowedMethod } = methodGuard;
@@ -71,6 +68,7 @@ for (const [key, value] of Object.entries(mergedEnv)) {
 // '@'` on the `@import "tailwindcss"` line. Force NODE_ENV to track the run
 // mode, exactly like the `next` CLI does.
 process.env.NODE_ENV = dev ? "development" : "production";
+process.env.OMNIROUTE_INTERNAL_SCHEME = "http";
 
 const { dashboardPort } = runtimePorts;
 const hostname = process.env.HOST || "0.0.0.0";
@@ -119,7 +117,8 @@ async function prepareWithHeal() {
   try {
     await nextApp.prepare();
   } catch (error) {
-    const detail = error instanceof Error ? `${error.message}\n${error.stack ?? ""}` : String(error);
+    const detail =
+      error instanceof Error ? `${error.message}\n${error.stack ?? ""}` : String(error);
     if (!useTurbopack || !isTurbopackCacheCorruption(detail)) throw error;
     console.warn(
       "[Next] Turbopack dev cache looks corrupted (Windows mmap / os error 1455 — known upstream bug). Purging and retrying once…"

scripts/dev/standalone-server-ws.mjs

@@ -18,6 +18,7 @@ const { wrapRequestListenerWithHeadResponseGuard } = headResponseGuard;
 // listener Next binds to (so WS `upgrade` / request wrappers keep working over
 // TLS). Absent or misconfigured → null → identical plain-HTTP behavior as before.
 const tlsOptions = resolveTlsOptions(process.env);
+process.env.OMNIROUTE_INTERNAL_SCHEME = tlsOptions ? "https" : "http";
 if (tlsOptions) {
   console.log(`[omniroute][tls] HTTPS enabled — terminating TLS with cert=${tlsOptions.certPath}`);
 }

scripts/i18n/run-translation.mjs

@@ -25,7 +25,7 @@
  * Backend (configured via env, never committed):
  *   OMNIROUTE_TRANSLATION_API_URL     e.g. https://cloud.omniroute.dev/v1
  *   OMNIROUTE_TRANSLATION_API_KEY     bearer token (kept out of logs)
- *   OMNIROUTE_TRANSLATION_MODEL       e.g. cx/gpt-5.4-mini
+ *   OMNIROUTE_TRANSLATION_MODEL       e.g. cx/gpt-5.6-sol
  *   OMNIROUTE_TRANSLATION_TIMEOUT_MS  optional, default 60000
  *   OMNIROUTE_TRANSLATION_CONCURRENCY optional, default 4
  */

src/app/(dashboard)/dashboard/cli-code/components/CodexToolCard.tsx

@@ -8,6 +8,7 @@ import { useTranslations } from "next-intl";
 import ProviderIcon from "@/shared/components/ProviderIcon";
 import { normalizeCodexBaseUrl } from "@/shared/utils/codexBaseUrl";
 import { isApplyDisabled, isResetDisabled } from "./codexButtonState";
+import { CODEX_DEFAULT_MODELS } from "./codexToolOptions";
 
 export default function CodexToolCard({
   tool,
@@ -28,14 +29,7 @@ export default function CodexToolCard({
   const [message, setMessage] = useState(null);
   const [showInstallGuide, setShowInstallGuide] = useState(false);
   const [selectedApiKey, setSelectedApiKey] = useState("");
-  const [selectedModel, setSelectedModel] = useState("gpt-5.5");
-  const CODEX_DEFAULT_MODELS = [
-    "gpt-5.5",
-    "gpt-5.3-codex",
-    "gpt-5.4",
-    "gpt-5.1-codex-max",
-    "gpt-5.1-codex-mini",
-  ];
+  const [selectedModel, setSelectedModel] = useState("gpt-5.6-sol");
   const [modelMappings, setModelMappings] = useState<Record<string, string>>({});
   const [reasoningEffort, setReasoningEffort] = useState("xhigh");
   const [wireApi, setWireApi] = useState("chat");
@@ -595,7 +589,7 @@ openai_base_url = "${getEffectiveBaseUrl()}"
                     type="text"
                     value={selectedModel}
                     onChange={(e) => setSelectedModel(e.target.value)}
-                    placeholder="gpt-5.5"
+                    placeholder="gpt-5.6-sol"
                     className="flex-1 px-2 py-1.5 bg-surface rounded border border-border text-xs focus:outline-none focus:ring-1 focus:ring-primary/50"
                   />
                   {selectedModel && (
@@ -627,6 +621,8 @@ openai_base_url = "${getEffectiveBaseUrl()}"
                     <option value="medium">Medium</option>
                     <option value="high">High</option>
                     <option value="xhigh">XHigh</option>
+                    <option value="max">Max</option>
+                    <option value="ultra">Ultra</option>
                   </select>
                 </div>
 

src/app/(dashboard)/dashboard/cli-code/components/codexToolOptions.ts

@@ -0,0 +1,9 @@
+export const CODEX_DEFAULT_MODELS = [
+  "gpt-5.6-sol",
+  "gpt-5.6-terra",
+  "gpt-5.6-luna",
+  "gpt-5.5",
+  "gpt-5.3-codex",
+  "gpt-5.1-codex-max",
+  "gpt-5.1-codex-mini",
+] as const;

src/app/(dashboard)/dashboard/combos/page.tsx

@@ -4090,7 +4090,7 @@ function ComboFormModal({ isOpen, combo, onClose, onSave, activeProviders, combo
                         <input
                           type="text"
                           value={config.handoffModel ?? ""}
-                          placeholder="codex/gpt-5.4"
+                          placeholder="codex/gpt-5.6-sol"
                           onChange={(e) =>
                             setConfig({
                               ...config,

src/app/(dashboard)/dashboard/providers/[id]/providerPageHelpers.ts

@@ -698,6 +698,7 @@ export const CODEX_REASONING_STRENGTH_OPTIONS = [
   { value: "medium", label: "Medium" },
   { value: "high", label: "High" },
   { value: "xhigh", label: "XHigh" },
+  { value: "max", label: "Max" },
 ];
 
 export const CODEX_ACCOUNT_SERVICE_TIER_VALUES: CodexServiceTier[] = [

src/app/(dashboard)/dashboard/settings/components/CodexFastTierTab.tsx

@@ -13,7 +13,12 @@ type TierValue = "default" | "priority" | "flex";
 // Fast-eligible Codex models per OpenAI ~/.codex/models_cache.json (service_tiers: priority).
 // Other future Fast-eligible slugs can be added here without code changes once the user
 // opts them in via the checkbox UI.
-const CODEX_FAST_TIER_CATALOG: readonly string[] = ["gpt-5.5", "gpt-5.4"];
+const CODEX_FAST_TIER_CATALOG: readonly string[] = [
+  "gpt-5.6-sol",
+  "gpt-5.6-terra",
+  "gpt-5.6-luna",
+  "gpt-5.5",
+];
 
 export default function CodexFastTierTab() {
   const [enabled, setEnabled] = useState(false);

src/app/(dashboard)/dashboard/settings/components/ComboDefaultsTab.tsx

@@ -636,7 +636,7 @@ export default function ComboDefaultsTab() {
               label={translateOrFallback(t, "contextRelaySummaryModel", "Summary Model")}
               type="text"
               value={comboDefaults.handoffModel ?? ""}
-              placeholder="codex/gpt-5.4"
+              placeholder="codex/gpt-5.6-sol"
               onChange={(e) =>
                 setComboDefaults((prev) => ({
                   ...prev,

src/app/api/cli-tools/codex-settings/route.ts

@@ -260,7 +260,7 @@ export async function POST(request: Request) {
     parsed._root.model = model;
 
     if (reasoningEffort && reasoningEffort !== "none") {
-      // Optional: low, medium, high
+      // Optional Codex reasoning effort.
       parsed._root.model_reasoning_effort = reasoningEffort;
     } else {
       delete parsed._root.model_reasoning_effort;

src/app/api/providers/[id]/models/discovery/codex.ts

@@ -2,6 +2,13 @@ import {
   getCodexClientVersion,
   getCodexDefaultHeaders,
 } from "@omniroute/open-sse/config/codexClient.ts";
+import { isCodexDiscoveryModelExcluded } from "@/shared/services/codexDiscoveryPolicy";
+
+export {
+  CODEX_DISCOVERY_EXCLUDED_IDS,
+  CODEX_DISCOVERY_EXCLUDED_ID_PREFIXES,
+  isCodexDiscoveryModelExcluded,
+} from "@/shared/services/codexDiscoveryPolicy";
 
 export const CODEX_MODELS_URL = "https://chatgpt.com/backend-api/codex/models";
 export const CODEX_GITHUB_MODELS_URL =
@@ -213,8 +220,9 @@ function getFreshCodexGithubCatalogCache(
   now: number,
   cacheTtlMs: number
 ): CodexDiscoveryModel[] | null {
-  if (cacheTtlMs > 0 && codexGithubCatalogCache?.expiresAt > now) {
-    return codexGithubCatalogCache.models;
+  const cache = codexGithubCatalogCache;
+  if (cacheTtlMs > 0 && cache && cache.expiresAt > now) {
+    return cache.models;
   }
   return null;
 }
@@ -284,6 +292,11 @@ function localCatalogModelToCodexDiscoveryModel(
   };
 }
 
+/**
+ * Live/GitHub discovery is the source of truth for "what exists".
+ * Explicit filters (denylist / predicates) are the policy layer for "what we show".
+ * Do NOT reintroduce curated-only allowlisting as the default path (#6862 / #6859).
+ */
 export function mergeCodexLiveModelsWithLocalCatalog(
   liveModels: CodexDiscoveryModel[],
   localCatalogModels: CodexLocalCatalogModel[]
@@ -291,6 +304,7 @@ export function mergeCodexLiveModelsWithLocalCatalog(
   const merged = new Map<string, CodexDiscoveryModel>();
 
   for (const liveModel of liveModels) {
+    if (!liveModel?.id) continue;
     merged.set(liveModel.id, liveModel);
   }
 
@@ -304,6 +318,64 @@ export function mergeCodexLiveModelsWithLocalCatalog(
   return Array.from(merged.values());
 }
 
+/** Return true to KEEP the model. */
+export type CodexDiscoveryModelFilter = (model: CodexDiscoveryModel) => boolean;
+
+/**
+ * Apply policy filters after discovery merge. Default denylist runs first;
+ * extraFilters are additional keep-predicates (all must pass).
+ */
+export function applyCodexDiscoveryFilters(
+  models: CodexDiscoveryModel[],
+  extraFilters: readonly CodexDiscoveryModelFilter[] = []
+): CodexDiscoveryModel[] {
+  return models.filter((model) => {
+    if (isCodexDiscoveryModelExcluded(model)) return false;
+    return extraFilters.every((keep) => keep(model));
+  });
+}
+
+/** Convenience: merge live/local then apply default (+ optional) filters. */
+export function buildCodexDiscoveryCatalog(
+  remoteModels: CodexDiscoveryModel[],
+  localCatalogModels: CodexLocalCatalogModel[],
+  extraFilters: readonly CodexDiscoveryModelFilter[] = []
+): CodexDiscoveryModel[] {
+  return applyCodexDiscoveryFilters(
+    mergeCodexLiveModelsWithLocalCatalog(remoteModels, localCatalogModels),
+    extraFilters
+  );
+}
+
+export type CuratedCodexCatalogResult = {
+  models: CodexDiscoveryModel[];
+  candidateModels: CodexDiscoveryModel[];
+};
+
+/**
+ * Optional curated-only view (allowlist). NOT used by the default Codex
+ * discovery route — kept for diagnostics / explicit call sites only.
+ */
+export function reconcileCuratedCodexCatalog(
+  remoteModels: CodexDiscoveryModel[],
+  curatedModels: CodexLocalCatalogModel[]
+): CuratedCodexCatalogResult {
+  const remoteById = new Map(remoteModels.map((model) => [model.id, model]));
+  const curatedIds = new Set<string>();
+  const models: CodexDiscoveryModel[] = [];
+
+  for (const localModel of curatedModels) {
+    if (!localModel.id) continue;
+    curatedIds.add(localModel.id);
+    const normalizedLocal = localCatalogModelToCodexDiscoveryModel(localModel);
+    const remoteModel = remoteById.get(localModel.id);
+    models.push(remoteModel ? { ...remoteModel, ...normalizedLocal } : normalizedLocal);
+  }
+
+  const candidateModels = remoteModels.filter((model) => !curatedIds.has(model.id));
+  return { models, candidateModels };
+}
+
 export function enrichCodexModelsFromGithubCatalog(
   models: CodexDiscoveryModel[],
   githubCatalogModels: CodexDiscoveryModel[]

src/app/api/providers/[id]/models/route.ts

@@ -98,10 +98,10 @@ import {
   PROVIDER_MODELS_CONFIG,
 } from "./discovery/providerModelsConfig";
 import {
+  buildCodexDiscoveryCatalog,
   enrichCodexModelsFromGithubCatalog,
   fetchCodexDiscoveryModels,
   fetchCodexGithubCatalogModels,
-  mergeCodexLiveModelsWithLocalCatalog,
 } from "./discovery/codex";
 
 /**
@@ -118,6 +118,7 @@ export async function GET(
     // Check if we should exclude hidden models (used by MCP tools to prevent hidden model leaks)
     const { searchParams } = new URL(request.url);
     const excludeHidden = searchParams.get("excludeHidden") === "true";
+    const excludeCustom = searchParams.get("excludeCustom") === "true";
     const refresh = searchParams.get("refresh") === "true";
 
     const connection = await getProviderConnectionById(id);
@@ -223,15 +224,19 @@ export async function GET(
     // per-connection route (used by MCP list_models_catalog + the dashboard
     // import view) never did, so custom models were dropped on both the
     // discovery-success and local_catalog paths. Read them once here and fold
-    // them into every models response via buildResponse below (dedup by id).
+    // them into every user-facing models response via buildResponse below
+    // (dedup by id). Internal model-sync discovery opts out because these rows
+    // are a response projection, not provider-discovered models.
     let customModelsForProvider: Array<{ id: string; name?: string }> = [];
-    try {
-      const custom = await getCustomModels(provider);
-      if (Array.isArray(custom)) {
-        customModelsForProvider = custom as Array<{ id: string; name?: string }>;
+    if (!excludeCustom) {
+      try {
+        const custom = await getCustomModels(provider);
+        if (Array.isArray(custom)) {
+          customModelsForProvider = custom as Array<{ id: string; name?: string }>;
+        }
+      } catch {
+        // DB unavailable — proceed without custom models.
       }
-    } catch {
-      // DB unavailable — proceed without custom models.
     }
 
     const mergeCustomModels = (models: any[]) => {
@@ -393,7 +398,11 @@ export async function GET(
       });
     };
 
-    const buildApiDiscoveryResponse = async (models: any[], warning?: string) => {
+    const buildApiDiscoveryResponse = async (
+      models: any[],
+      warning?: string,
+      extraPayload: Record<string, unknown> = {}
+    ) => {
       const discoveredModels = await persistDiscoveredModels(provider, connectionId, models);
       if (discoveredModels.length > 0) {
         return buildResponse({
@@ -402,6 +411,7 @@ export async function GET(
           models,
           source: "api",
           ...(warning ? { warning } : {}),
+          ...extraPayload,
         });
       }
 
@@ -1721,11 +1731,42 @@ export async function GET(
     }
 
     if (provider === "codex") {
-      const cachedResponse = maybeReturnCachedDiscovery();
-      if (cachedResponse) return cachedResponse;
+      // Auto-merge live/GitHub/local (future-proof discovery), then apply explicit
+      // denylist filters (e.g. drop GPT-5.4 family). Do not gate remote-only IDs.
+      const staticCodexCatalog = mergeLocalCatalogModels(
+        getModelsByProviderId("codex") || [],
+        getStaticModelsForProvider("codex") || []
+      );
+      const finalizeCodexCatalog = (remoteModels: typeof cachedDiscoveryModels) =>
+        buildCodexDiscoveryCatalog(remoteModels, staticCodexCatalog);
+      const cachedCatalogModels = finalizeCodexCatalog(cachedDiscoveryModels);
+      const cachedIdsMatchFinalCatalog =
+        cachedDiscoveryModels.length === cachedCatalogModels.length &&
+        cachedDiscoveryModels.every((model, index) => model.id === cachedCatalogModels[index]?.id);
+      const persistFilteredCacheIfNeeded = async () => {
+        if (cachedIdsMatchFinalCatalog) return;
+        await persistDiscoveredModels(provider, connectionId, cachedCatalogModels);
+      };
 
-      const autoFetchDisabledResponse = maybeReturnAutoFetchDisabled();
-      if (autoFetchDisabledResponse) return autoFetchDisabledResponse;
+      if (!refresh && cachedDiscoveryModels.length > 0) {
+        await persistFilteredCacheIfNeeded();
+        return buildResponse({
+          provider,
+          connectionId,
+          models: cachedCatalogModels,
+          source: "cache",
+        });
+      }
+
+      if (!refresh && !autoFetchModels) {
+        return buildResponse({
+          provider,
+          connectionId,
+          models: finalizeCodexCatalog([]),
+          source: "local_catalog",
+          warning: "Auto-fetch disabled — using local catalog",
+        });
+      }
 
       const liveModels = await fetchCodexDiscoveryModels({
         accessToken: accessToken || null,
@@ -1747,38 +1788,35 @@ export async function GET(
             ...init,
           }),
       });
-      const staticCodexCatalog = mergeLocalCatalogModels(
-        getModelsByProviderId("codex") || [],
-        getStaticModelsForProvider("codex") || []
-      );
-
       if (liveModels && liveModels.length > 0) {
         const enrichedLiveModels =
           githubCatalogModels && githubCatalogModels.length > 0
             ? enrichCodexModelsFromGithubCatalog(liveModels, githubCatalogModels)
             : liveModels;
-        return buildApiDiscoveryResponse(
-          mergeCodexLiveModelsWithLocalCatalog(enrichedLiveModels, staticCodexCatalog)
-        );
+        return buildApiDiscoveryResponse(finalizeCodexCatalog(enrichedLiveModels));
       }
 
       if (githubCatalogModels && githubCatalogModels.length > 0) {
         return buildApiDiscoveryResponse(
-          mergeCodexLiveModelsWithLocalCatalog(githubCatalogModels, staticCodexCatalog),
+          finalizeCodexCatalog(githubCatalogModels),
           "Codex live catalog unavailable — using GitHub model catalog"
         );
       }
 
-      const fallback = buildDiscoveryFallbackResponse({
-        cacheWarning: "Codex live catalog unavailable — using cached catalog",
-        localWarning: "Codex live and GitHub catalogs unavailable — using local catalog",
-        localIntentional: true,
-      });
-      if (fallback) return fallback;
+      if (cachedDiscoveryModels.length > 0) {
+        await persistFilteredCacheIfNeeded();
+        return buildResponse({
+          provider,
+          connectionId,
+          models: cachedCatalogModels,
+          source: "cache",
+          warning: "Codex live catalog unavailable — using cached catalog",
+        });
+      }
       return buildResponse({
         provider,
         connectionId,
-        models: [],
+        models: finalizeCodexCatalog([]),
         source: "local_catalog",
         intentional: true,
         warning: "Codex live and GitHub catalogs unavailable — using local catalog",

src/app/api/providers/[id]/sync-models/route.ts

@@ -10,6 +10,8 @@ import { saveCallLog } from "@/lib/usage/callLogs";
 import { isAuthenticated } from "@/shared/utils/apiAuth";
 import {
   buildModelSyncInternalHeaders,
+  fetchModelSyncInternal,
+  getModelSyncInternalBaseUrl,
   isModelSyncInternalRequest,
 } from "@/shared/services/modelSyncScheduler";
 import { autoSyncCodexProfilesFromLiveCatalog } from "@/lib/cli-helper/codexProfileAutoSync";
@@ -180,7 +182,7 @@ export type EnsureReadyOptions = {
 export async function ensureLoopbackServerReady(opts: EnsureReadyOptions = {}): Promise<void> {
   if (__loopbackReadyPromise != null) return __loopbackReadyPromise;
   __loopbackReadyPromise = (async () => {
-    const f = opts.fetch ?? fetch;
+    const f = opts.fetch ?? fetchModelSyncInternal;
     const maxWaitMs = opts.maxWaitMs ?? 30_000;
     const pollMs = opts.pollMs ?? 250;
     const deadline = Date.now() + maxWaitMs;
@@ -191,10 +193,10 @@ export async function ensureLoopbackServerReady(opts: EnsureReadyOptions = {}):
         // readiness — we only care that the dispatcher succeeds (no
         // ECONNREFUSED). Using a synthetic connection id so no real DB lookup
         // is needed; the 404 is sufficient proof the server is dispatching.
-        const probePort = process.env.OMNIROUTE_PORT || process.env.PORT || "20128";
         const res = await f(
-          `http://127.0.0.1:${probePort}/api/providers/__readiness_probe__/models`,
+          `${getModelSyncInternalBaseUrl()}/api/providers/__readiness_probe__/models`,
           {
+            redirect: "error",
             signal: AbortSignal.timeout(2_000),
           }
         );
@@ -264,7 +266,7 @@ export async function selfFetchWithRetry(
   url: string,
   opts: SelfFetchWithRetryOptions = {}
 ): Promise<Response> {
-  const f = opts.fetch ?? fetch;
+  const f = opts.fetch ?? fetchModelSyncInternal;
   // Reduced from 5 to 3: the readiness gate now handles the boot race.
   // Retries here are only for transient failures after server is confirmed up.
   const maxRetries = opts.maxRetries ?? 3;
@@ -321,30 +323,20 @@ export async function selfFetchWithRetry(
 // ---------------------------------------------------------------------------
 
 async function fetchProviderModelsForSync(request: Request, connectionId: string) {
-  // Construct a safe localhost URL from the incoming request's origin.
-  // The route only accepts authenticated or internal-scheduler requests,
-  // and the path is hardcoded — no user-controlled URL components reach fetch.
-  // Always use 127.0.0.1 (IPv4) — never "localhost" which may resolve to ::1
-  // (IPv6) in containers, causing TypeError: fetch failed even when the HTTP
-  // server is bound only to 0.0.0.0 (IPv4 only).
-  const SAFE_HOSTS = new Set(["localhost", "127.0.0.1", "0.0.0.0", "::1"]);
-  const incomingUrl = new URL(request.url);
-  const loopbackPort =
-    SAFE_HOSTS.has(incomingUrl.hostname) && incomingUrl.port
-      ? incomingUrl.port
-      : process.env.PORT || "20128";
-  const safeOrigin = `http://127.0.0.1:${loopbackPort}`;
-  const modelsPath = `/api/providers/${encodeURIComponent(connectionId)}/models?refresh=true`;
+  const safeOrigin = getModelSyncInternalBaseUrl();
+  const modelsPath =
+    `/api/providers/${encodeURIComponent(connectionId)}/models` +
+    "?refresh=true&excludeCustom=true";
   const headers = {
     cookie: request.headers.get("cookie") || "",
     ...buildModelSyncInternalHeaders(),
   };
 
-  const targetUrl = new URL(modelsPath, safeOrigin).href;
+  const targetUrl = `${safeOrigin}${modelsPath}`;
 
   // Wrap fetch so it forwards the required headers on every retry attempt.
   const fetchWithHeaders: typeof fetch = (input, init) =>
-    fetch(input as string, { ...init, headers });
+    fetchModelSyncInternal(input, { ...init, headers, redirect: "error" });
 
   return selfFetchWithRetry(targetUrl, {
     fetch: fetchWithHeaders,
@@ -376,9 +368,12 @@ async function fetchProviderModelsForSync(request: Request, connectionId: string
 export async function POST(request: Request, { params }: { params: Promise<{ id: string }> }) {
   const start = Date.now();
   const { id } = await params;
+  const requestUrl = new URL(request.url);
   const mode = (
-    new URL(request.url).searchParams.get("mode") === "import" ? "merge" : "sync"
+    requestUrl.searchParams.get("mode") === "import" ? "merge" : "sync"
   ) as ManagedModelImportMode;
+  // quiet=1: boot revalidation path — skip chatty ModelSync console lines
+  const quiet = requestUrl.searchParams.get("quiet") === "1";
   let logProvider = "unknown";
   let channelLabel: string | null = null;
 
@@ -523,7 +518,7 @@ export async function POST(request: Request, { params }: { params: Promise<{ id:
     const updatedCount = importedChanges.updated;
     const shouldLog = modelChanges.total > 0 || customModelChanges.total > 0;
 
-    if (shouldLog) {
+    if (shouldLog && !quiet) {
       void autoSyncCodexProfilesFromLiveCatalog(request, `model-sync:${logProvider}`)
         .then((syncResult) => {
           if (syncResult.ok) {
@@ -561,6 +556,14 @@ export async function POST(request: Request, { params }: { params: Promise<{ id:
             err?.message || err
           );
         });
+    } else if (shouldLog && quiet) {
+      // Still update profiles; suppress console noise from boot revalidation.
+      void autoSyncCodexProfilesFromLiveCatalog(request, `model-sync:${logProvider}`).catch(
+        () => undefined
+      );
+      void autoSyncClaudeProfilesFromLiveCatalog(request, `model-sync:${logProvider}`).catch(
+        () => undefined
+      );
     }
 
     if (shouldLog) {

src/app/api/providers/route.ts

@@ -34,6 +34,7 @@ import { isManagedProviderConnectionId } from "@/lib/providers/catalog";
 import { isApiKeyRevealEnabled, maskStoredApiKey } from "@/lib/apiKeyExposure";
 import {
   buildModelSyncInternalHeaders,
+  fetchModelSyncInternal,
   getModelSyncInternalBaseUrl,
 } from "@/shared/services/modelSyncScheduler";
 
@@ -194,7 +195,11 @@ export async function POST(request: Request) {
       };
       const syncUrl = `${internalOrigin}/api/providers/${encodeURIComponent(newConnection.id)}/sync-models?mode=import`;
       // Intentionally not awaited: this is async/non-blocking work.
-      void fetch(syncUrl, { method: "POST", headers: syncHeaders })
+      void fetchModelSyncInternal(syncUrl, {
+        method: "POST",
+        headers: syncHeaders,
+        redirect: "error",
+      })
         .then((syncRes) => {
           if (!syncRes.ok) {
             console.log(`[providers] Auto-sync failed for ${newConnection.id}: ${syncRes.status}`);

src/app/api/sync/initialize/route.ts

@@ -1,7 +1,6 @@
 import { NextResponse } from "next/server";
 import initializeCloudSync from "@/shared/services/initializeCloudSync";
 import { startModelSyncScheduler } from "@/shared/services/modelSyncScheduler";
-import { resolveOmniRouteBaseUrl } from "@/shared/utils/resolveOmniRouteBaseUrl";
 
 let syncInitialized = false;
 let modelSyncInitialized = false;
@@ -20,8 +19,7 @@ export async function POST(request) {
 
     // (#488) Start model auto-sync scheduler (24h, configurable via MODEL_SYNC_INTERVAL_HOURS)
     if (!modelSyncInitialized) {
-      const origin = request.headers.get("origin") || resolveOmniRouteBaseUrl();
-      startModelSyncScheduler(origin);
+      startModelSyncScheduler();
       modelSyncInitialized = true;
     }
 

src/app/api/v1/images/edits/route.ts

@@ -157,7 +157,7 @@ async function postHandler(request: Request, context) {
     return errorResponse(HTTP_STATUS.BAD_REQUEST, "Missing required field: image");
   }
 
-  const fullModel = model || "cgpt-web/gpt-5.3-instant";
+  const fullModel = model || "cgpt-web/gpt-5.5";
 
   const policy = await enforceApiKeyPolicy(request, fullModel);
   if (policy.rejection) return policy.rejection;

src/app/api/v1/models/catalog.ts

@@ -12,7 +12,10 @@ import {
 import { extractAliasBackedModels } from "./aliasBackedModels";
 import { appendNoThinkingVariants } from "@omniroute/open-sse/utils/noThinkingAlias";
 import { getAllEmbeddingModels } from "@omniroute/open-sse/config/embeddingRegistry";
-import { getAllImageModels, isRegisteredImageModel } from "@omniroute/open-sse/config/imageRegistry";
+import {
+  getAllImageModels,
+  isRegisteredImageModel,
+} from "@omniroute/open-sse/config/imageRegistry";
 import { getAllRerankModels } from "@omniroute/open-sse/config/rerankRegistry";
 import { getAllAudioModels } from "@omniroute/open-sse/config/audioRegistry";
 import { getAllModerationModels } from "@omniroute/open-sse/config/moderationRegistry";
@@ -78,6 +81,7 @@ import { getVisionCapabilityFields, getCustomVisionCapabilityFields } from "./ca
 import { FALLBACK_ALIAS_TO_PROVIDER, buildAliasMaps } from "./catalogProviderMaps";
 import { getModelCatalogAuthRejection, isCodexModelCatalogClient } from "./catalogRequest";
 import { isFreeModel, providerHasFreeModels } from "@/shared/utils/freeModels";
+import { isCodexDiscoveryModelExcluded } from "@/shared/services/codexDiscoveryPolicy";
 
 // Public API of this module is preserved after the catalog helper extraction:
 // `isVisionModelId` (vision-detection-consistency.test.ts) and
@@ -760,7 +764,8 @@ async function buildUnifiedModelsResponseCore(
         if (!providerSupportsModel(canonicalProviderId, model.id)) continue;
         const aliasId = `${alias}/${model.id}`;
         if (getModelIsHidden(canonicalProviderId, model.id)) continue;
-        if (shouldHidePaid(canonicalProviderId, model.id, (model as { pricing?: unknown }).pricing)) continue;
+        if (shouldHidePaid(canonicalProviderId, model.id, (model as { pricing?: unknown }).pricing))
+          continue;
 
         const visionFields =
           getVisionCapabilityFields(aliasId) || getVisionCapabilityFields(model.id);
@@ -848,6 +853,9 @@ async function buildUnifiedModelsResponseCore(
 
         for (const sm of syncedModels) {
           if (!providerSupportsModel(canonicalProviderId, sm.id)) continue;
+          if (canonicalProviderId === "codex" && isCodexDiscoveryModelExcluded(sm)) {
+            continue;
+          }
           if (getModelIsHidden(providerId, sm.id)) continue;
           // #6457: some upstream discovery catalogs (e.g. HuggingFace's live
           // `/v1/models`) return image/diffusion models with no modality info,
@@ -866,7 +874,8 @@ async function buildUnifiedModelsResponseCore(
           // the FREE_MODEL_IDS_BY_PROVIDER catalog — providers with a curated
           // free roster show only those; providers with none fall through to
           // hide-all via providerHasFreeModels() === false.
-          if (shouldHidePaid(canonicalProviderId, sm.id, (sm as { pricing?: unknown }).pricing)) continue;
+          if (shouldHidePaid(canonicalProviderId, sm.id, (sm as { pricing?: unknown }).pricing))
+            continue;
 
           const registryEntry = REGISTRY[providerId];
           const displayModelId =
@@ -1227,7 +1236,10 @@ async function buildUnifiedModelsResponseCore(
           // #6328: apply hidePaidModels to user-defined custom rows too.
           // Custom entries do not carry pricing, so shouldHidePaid() decides
           // via FREE_MODEL_IDS_BY_PROVIDER — matches synced/PROVIDER_MODELS.
-          if (shouldHidePaid(canonicalProviderId, modelId, (model as { pricing?: unknown }).pricing)) continue;
+          if (
+            shouldHidePaid(canonicalProviderId, modelId, (model as { pricing?: unknown }).pricing)
+          )
+            continue;
           // noAuth providers have no connection rows; keep auth providers gated. (#2798/#3200)
           const isNoAuthProvider = isNoAuthProviderKey(canonicalProviderId, providerId, alias);
           if (
@@ -1428,7 +1440,8 @@ async function buildUnifiedModelsResponseCore(
         // #6328: apply hidePaidModels to managed-fallback rows too. Compatible
         // provider fallbacks lack pricing; shouldHidePaid() decides via the
         // FREE_MODEL_IDS_BY_PROVIDER catalog tier.
-        if (shouldHidePaid(canonicalProviderId, modelId, (model as { pricing?: unknown }).pricing)) continue;
+        if (shouldHidePaid(canonicalProviderId, modelId, (model as { pricing?: unknown }).pricing))
+          continue;
         if (!hasEligibleConnectionForModel([conn], modelId)) continue;
 
         const aliasId = `${alias}/${modelId}`;

src/lib/cli-helper/claudeProfileAutoSync.ts

@@ -1,6 +1,9 @@
 import path from "node:path";
 import { ensureCliConfigWriteAllowed, getCliConfigPaths } from "../../shared/services/cliRuntime";
-import { getModelSyncInternalBaseUrl } from "../../shared/services/modelSyncScheduler";
+import {
+  fetchModelSyncInternal,
+  getModelSyncInternalBaseUrl,
+} from "../../shared/services/modelSyncScheduler";
 import { isFeatureFlagEnabled } from "../../shared/utils/featureFlags";
 
 type SyncResult =
@@ -47,8 +50,9 @@ export async function autoSyncClaudeProfilesFromLiveCatalog(
   }
 
   const internalBase = getModelSyncInternalBaseUrl().replace(/\/$/, "");
-  const res = await fetch(`${internalBase}/v1/models`, {
+  const res = await fetchModelSyncInternal(`${internalBase}/v1/models`, {
     headers: forwardAuthHeaders(request),
+    redirect: "error",
     signal: AbortSignal.timeout(10_000),
   });
 

src/lib/cli-helper/codexProfileAutoSync.ts

@@ -1,6 +1,9 @@
 import path from "node:path";
 import { ensureCliConfigWriteAllowed, getCliConfigPaths } from "../../shared/services/cliRuntime";
-import { getModelSyncInternalBaseUrl } from "../../shared/services/modelSyncScheduler";
+import {
+  fetchModelSyncInternal,
+  getModelSyncInternalBaseUrl,
+} from "../../shared/services/modelSyncScheduler";
 import { isFeatureFlagEnabled } from "../../shared/utils/featureFlags";
 
 type SyncResult =
@@ -47,8 +50,9 @@ export async function autoSyncCodexProfilesFromLiveCatalog(
   }
 
   const baseUrl = getModelSyncInternalBaseUrl().replace(/\/$/, "");
-  const res = await fetch(`${baseUrl}/v1/models`, {
+  const res = await fetchModelSyncInternal(`${baseUrl}/v1/models`, {
     headers: forwardAuthHeaders(request),
+    redirect: "error",
     signal: AbortSignal.timeout(10_000),
   });
 

src/lib/db/settings.ts

@@ -190,6 +190,17 @@ export async function getSettings() {
 }
 
 export async function updateSettings(updates: Record<string, unknown>) {
+  // Detect first-time setup completion before we overwrite settings.
+  let setupJustCompleted = false;
+  if (updates.setupComplete === true) {
+    try {
+      const prev = await getSettings();
+      setupJustCompleted = prev.setupComplete !== true;
+    } catch {
+      setupJustCompleted = true;
+    }
+  }
+
   const db = getDbInstance();
   const insert = db.prepare(
     "INSERT OR REPLACE INTO key_value (namespace, key, value) VALUES ('settings', ?, ?)"
@@ -221,6 +232,17 @@ export async function updateSettings(updates: Record<string, unknown>) {
     );
   }
 
+  // Onboarding / setup finished → one-shot Codex catalog revalidation (init case).
+  if (setupJustCompleted) {
+    void import("@/shared/services/codexCatalogRevalidation")
+      .then(({ scheduleCodexCatalogRevalidationAfterInit }) => {
+        scheduleCodexCatalogRevalidationAfterInit();
+      })
+      .catch(() => {
+        // non-fatal
+      });
+  }
+
   return nextSettings;
 }
 

src/lib/modelMetadataRegistry.ts

@@ -12,7 +12,10 @@ import {
 import { AI_PROVIDERS } from "@/shared/constants/providers";
 import { PROVIDER_ID_TO_ALIAS, PROVIDER_MODELS } from "@/shared/constants/models";
 import { getSyncStatus, getSyncedCapability } from "@/lib/modelsDevSync";
-import { CANONICAL_EFFORT_VALUES } from "@/shared/reasoning/effortStandardization";
+import {
+  CANONICAL_EFFORT_VALUES,
+  extendCodexGpt56EffortValues,
+} from "@/shared/reasoning/effortStandardization";
 
 const MODEL_METADATA_SCHEMA_VERSION = "model-metadata-v1";
 
@@ -296,7 +299,13 @@ export function enrichCatalogModelEntry<T extends JsonRecord>(
           thinking: metadata.capabilities.supportsThinking,
           supportsThinking: metadata.capabilities.supportsThinking,
           ...(metadata.capabilities.supportsThinking
-            ? { effort_tiers: [...CANONICAL_EFFORT_VALUES] }
+            ? {
+                effort_tiers: extendCodexGpt56EffortValues(
+                  metadata.provider,
+                  metadata.model,
+                  CANONICAL_EFFORT_VALUES
+                ),
+              }
             : {}),
         }
       : {}),

src/lib/providers/codexFastTier.ts

@@ -13,7 +13,12 @@ function asRecord(value: unknown): JsonRecord {
 export type CodexFastTierValue = CodexServiceTier;
 export type CodexGlobalServiceMode = "none" | CodexServiceTier;
 
-export const CODEX_FAST_TIER_DEFAULT_SUPPORTED_MODELS: readonly string[] = ["gpt-5.5", "gpt-5.4"];
+export const CODEX_FAST_TIER_DEFAULT_SUPPORTED_MODELS: readonly string[] = [
+  "gpt-5.6-sol",
+  "gpt-5.6-terra",
+  "gpt-5.6-luna",
+  "gpt-5.5",
+];
 
 export interface CodexGlobalFastServiceTierResolved {
   enabled: boolean;
@@ -30,7 +35,8 @@ export interface CodexGlobalFastServiceTierResolved {
  *
  * Defaults when fields are absent on an enabled config:
  *  - tier            = "priority"  (back-compat: PR #2440 only injected priority)
- *  - supportedModels = ["gpt-5.5", "gpt-5.4"] (OpenAI Fast-eligible per models_cache.json)
+ *  - supportedModels = CODEX_FAST_TIER_DEFAULT_SUPPORTED_MODELS
+ *    (OpenAI Fast-eligible per models_cache.json)
  */
 export function resolveCodexGlobalFastServiceTier(
   settings: unknown

src/lib/providers/requestDefaults.ts

@@ -5,7 +5,14 @@ import { normalizeExcludedModelPatterns } from "@/domain/connectionModelRules";
 import { normalizeRoutingTags } from "@/domain/tagRouter";
 import { normalizeOpenRouterPreset } from "@/shared/constants/openRouterPreset";
 
-export const CODEX_REASONING_EFFORT_VALUES = ["none", "low", "medium", "high", "xhigh"] as const;
+export const CODEX_REASONING_EFFORT_VALUES = [
+  "none",
+  "low",
+  "medium",
+  "high",
+  "xhigh",
+  "max",
+] as const;
 
 export type CodexReasoningEffort = (typeof CODEX_REASONING_EFFORT_VALUES)[number];
 

src/lib/usage/costCalculator.ts

@@ -79,7 +79,7 @@ function normalizeServiceTier(value: unknown): string {
 }
 
 function stripCodexEffortSuffix(model: string): string {
-  return model.replace(/-(?:xhigh|high|medium|low|none)$/i, "");
+  return model.replace(/-(?:ultra|max|xhigh|high|medium|low|none)$/i, "");
 }
 
 export function getCodexFastCostMultiplier(
@@ -101,6 +101,12 @@ export function getCodexFastCostMultiplier(
 
   const modelKey = stripCodexEffortSuffix(normalizeModelName(String(model || "")).toLowerCase());
   const compactModelKey = modelKey.replace(/-/g, "");
+  if (
+    /^gpt-5\.6-(?:sol|terra|luna)$/.test(modelKey) ||
+    /^gpt5\.6(?:sol|terra|luna)$/.test(compactModelKey)
+  ) {
+    return 1.5;
+  }
   if (modelKey === "gpt-5.5" || compactModelKey === "gpt5.5") return 2.5;
   if (modelKey === "gpt-5.4" || compactModelKey === "gpt5.4") return 2;
   return 1;
@@ -237,7 +243,10 @@ export function computeAudioCost(
   }
   const characters = toNumber(usage.characters, 0);
   if (characters > 0) {
-    const perChar = toNumber(pricing.input_cost_per_character ?? pricing.output_cost_per_character, 0);
+    const perChar = toNumber(
+      pricing.input_cost_per_character ?? pricing.output_cost_per_character,
+      0
+    );
     // Round to 10 decimals to drop binary-FP artifacts (e.g. 0.000015 * 1000).
     if (perChar > 0) return Math.round(perChar * characters * 1e10) / 1e10;
   }

src/lib/vscode/reasoningMetadata.ts

@@ -1,6 +1,7 @@
 import { supportsXHighEffort } from "@omniroute/open-sse/config/providerModels";
 import { parseModel } from "@omniroute/open-sse/services/model";
 import { stripVscodeServiceTierVariantModelId } from "@/lib/vscode/serviceTierVariants";
+import { extendCodexGpt56EffortValues } from "@/shared/reasoning/effortStandardization";
 
 export type VscodeCatalogModel = {
   id?: string;
@@ -15,9 +16,10 @@ export type VscodeCatalogModel = {
   default_reasoning_effort?: string;
 };
 
-const EFFORT_SUFFIX_PATTERN = /-(xhigh|high|medium|low|none)$/i;
+const STANDARD_EFFORT_SUFFIX_PATTERN = /-(xhigh|high|medium|low|none)$/i;
+const GPT_5_6_EXTENDED_EFFORT_SUFFIX_PATTERN = /^(.*gpt-5\.6-(?:sol|terra|luna))-(max|ultra)$/i;
 const DEFAULT_REASONING_EFFORT = "none";
-const KNOWN_REASONING_EFFORTS = new Set(["none", "low", "medium", "high", "xhigh"]);
+const KNOWN_REASONING_EFFORTS = new Set(["none", "low", "medium", "high", "xhigh", "max", "ultra"]);
 
 export type VscodeModelConfigSchema = {
   type: "object";
@@ -38,6 +40,20 @@ export function getCatalogModelName(model: VscodeCatalogModel) {
   return stripVscodeServiceTierVariantModelId(model.id || model.name || model.root || "");
 }
 
+function matchReasoningEffortSuffix(modelId: string) {
+  const extendedMatch = modelId.match(GPT_5_6_EXTENDED_EFFORT_SUFFIX_PATTERN);
+  if (extendedMatch?.[1] && extendedMatch[2]) {
+    return { baseModelId: extendedMatch[1], effort: extendedMatch[2].toLowerCase() };
+  }
+
+  const standardMatch = modelId.match(STANDARD_EFFORT_SUFFIX_PATTERN);
+  if (!standardMatch?.[1]) return undefined;
+  return {
+    baseModelId: modelId.slice(0, -standardMatch[0].length),
+    effort: standardMatch[1].toLowerCase(),
+  };
+}
+
 function normalizeReasoningEffortValue(value: string) {
   const normalized = value
     .trim()
@@ -104,7 +120,7 @@ export function getReasoningEffortValues(model: VscodeCatalogModel) {
     values.push("xhigh");
   }
 
-  return values;
+  return extendCodexGpt56EffortValues(providerId, providerModelId, values);
 }
 
 export function formatReasoningEffortLabel(level: string) {
@@ -123,7 +139,11 @@ function describeReasoningEffort(level: string) {
     case "high":
       return "Uses an extended amount of reasoning.";
     case "xhigh":
+      return "Uses extra-high reasoning effort.";
+    case "max":
       return "Uses the maximum available reasoning effort.";
+    case "ultra":
+      return "Uses the Ultra reasoning preset.";
     default:
       return `Uses ${formatReasoningEffortLabel(level)} reasoning effort.`;
   }
@@ -138,11 +158,10 @@ export function inferSelectedReasoningEffort(
   supportedValues?: string[]
 ) {
   const modelId = getCatalogModelName(model);
-  const match = modelId.match(EFFORT_SUFFIX_PATTERN);
+  const match = matchReasoningEffortSuffix(modelId);
   if (!match) return undefined;
 
-  const selected = match[1]?.toLowerCase();
-  if (!selected) return undefined;
+  const selected = match.effort;
   if (
     Array.isArray(supportedValues) &&
     supportedValues.length > 0 &&
@@ -155,7 +174,7 @@ export function inferSelectedReasoningEffort(
 }
 
 export function getReasoningVariantBaseModelId(modelId: string) {
-  return modelId.replace(EFFORT_SUFFIX_PATTERN, "");
+  return matchReasoningEffortSuffix(modelId)?.baseModelId || modelId;
 }
 
 export function getDefaultReasoningEffort(model: VscodeCatalogModel, supportedValues?: string[]) {

src/shared/constants/clientIdentityProfiles.ts

@@ -30,7 +30,7 @@ const CLAUDE_CLI_PROFILE: ClientIdentityProfile = Object.freeze({
   id: "claude-cli",
   label: "Claude CLI",
   headers: Object.freeze({
-    "User-Agent": "claude-cli/2.1.195 (external, cli)",
+    "User-Agent": "claude-cli/2.1.207 (external, cli)",
     "X-App": "cli",
   }),
 });
@@ -39,7 +39,7 @@ const CODEX_CLI_PROFILE: ClientIdentityProfile = Object.freeze({
   id: "codex-cli",
   label: "Codex CLI",
   headers: Object.freeze({
-    "User-Agent": "codex_cli_rs/0.136.0",
+    "User-Agent": "codex_cli_rs/0.144.1",
     originator: "codex_cli_rs",
   }),
 });
@@ -61,9 +61,7 @@ export const CLIENT_IDENTITY_PROFILES: Readonly<Record<string, ClientIdentityPro
     "gemini-cli": GEMINI_CLI_PROFILE,
   });
 
-export const CLIENT_IDENTITY_PROFILE_IDS: readonly string[] = Object.keys(
-  CLIENT_IDENTITY_PROFILES
-);
+export const CLIENT_IDENTITY_PROFILE_IDS: readonly string[] = Object.keys(CLIENT_IDENTITY_PROFILES);
 
 export const CLIENT_IDENTITY_PROFILE_OPTIONS: ReadonlyArray<{ value: string; label: string }> =
   CLIENT_IDENTITY_PROFILE_IDS.map((id) => ({
@@ -72,7 +70,10 @@ export const CLIENT_IDENTITY_PROFILE_OPTIONS: ReadonlyArray<{ value: string; lab
   }));
 
 export function isClientIdentityProfileId(value: unknown): value is string {
-  return typeof value === "string" && Object.prototype.hasOwnProperty.call(CLIENT_IDENTITY_PROFILES, value);
+  return (
+    typeof value === "string" &&
+    Object.prototype.hasOwnProperty.call(CLIENT_IDENTITY_PROFILES, value)
+  );
 }
 
 /**

src/shared/constants/modelSpecs.ts

@@ -65,7 +65,32 @@ const AUTHORITATIVE_PROVIDER_CONTEXT_WINDOWS = new Map<string, number>([
   ["zenmux/z-ai/glm-5.2-free", 1000000],
 ]);
 
+const GPT_5_6_MODEL_SPEC = {
+  maxOutputTokens: 128000,
+  contextWindow: 1050000,
+  supportsThinking: true,
+  supportsTools: true,
+  supportsVision: true,
+} satisfies ModelSpec;
+
 export const MODEL_SPECS: Record<string, ModelSpec> = {
+  "gpt-5.6": {
+    ...GPT_5_6_MODEL_SPEC,
+    aliases: ["openai/gpt-5.6"],
+  },
+  "gpt-5.6-sol": {
+    ...GPT_5_6_MODEL_SPEC,
+    aliases: ["openai/gpt-5.6-sol"],
+  },
+  "gpt-5.6-terra": {
+    ...GPT_5_6_MODEL_SPEC,
+    aliases: ["openai/gpt-5.6-terra"],
+  },
+  "gpt-5.6-luna": {
+    ...GPT_5_6_MODEL_SPEC,
+    aliases: ["openai/gpt-5.6-luna"],
+  },
+
   "gpt-5.5": {
     maxOutputTokens: 128000,
     contextWindow: 1050000,

src/shared/constants/pricing/frontier-labs.ts

@@ -4,6 +4,9 @@
  */
 import {
   GPT_5_5_PRICING,
+  GPT_5_6_LUNA_PRICING,
+  GPT_5_6_SOL_PRICING,
+  GPT_5_6_TERRA_PRICING,
   CLAUDE_FABLE_5_PRICING,
   CLAUDE_OPUS_4_PRICING,
   CLAUDE_SONNET_4_PRICING,
@@ -14,13 +17,16 @@ import {
 
 export const DEFAULT_PRICING_FRONTIER = {
   openai: {
+    "gpt-5.6": GPT_5_6_SOL_PRICING,
+    "gpt-5.6-sol": GPT_5_6_SOL_PRICING,
+    "gpt-5.6-terra": GPT_5_6_TERRA_PRICING,
+    "gpt-5.6-luna": GPT_5_6_LUNA_PRICING,
     "gpt-5.5": GPT_5_5_PRICING,
     // The -pro tier mirrors its base family pricing until OpenAI publishes a
     // distinct pro rate; without these rows the openai provider's gpt-5.x-pro
     // models (in the registry) resolved to $0 and tripped the catalog pricing gate.
     "gpt-5.5-pro": GPT_5_5_PRICING,
-    // gpt-5.4 family (public API tier; mirrors the codex 5.4 tier for the
-    // base/mini, with a lower nano tier). Without these rows the openai
+    // gpt-5.4 family (public API tier, with a lower nano tier). Without these rows the openai
     // provider's gpt-5.4* models resolved to $0.
     "gpt-5.4": {
       input: 5.0,

src/shared/constants/pricing/oauth-subscriptions.ts

@@ -2,7 +2,13 @@
  * Pricing data — oauth-subscriptions family (OAuth / IDE subscription aliases (cc, codex, antigravity, copilot, kiro…)).
  * Pure data; merged by default-pricing.ts via spread (god-file decomposition; semantic split).
  */
-import { GPT_5_3_CODEX_PRICING, GPT_5_5_PRICING } from "./shared-tiers";
+import {
+  GPT_5_3_CODEX_PRICING,
+  GPT_5_5_PRICING,
+  GPT_5_6_LUNA_PRICING,
+  GPT_5_6_SOL_PRICING,
+  GPT_5_6_TERRA_PRICING,
+} from "./shared-tiers";
 
 export const DEFAULT_PRICING_OAUTH = {
   cc: {
@@ -72,6 +78,28 @@ export const DEFAULT_PRICING_OAUTH = {
   },
   cx: {
     "codex-auto-review": GPT_5_5_PRICING,
+    // Codex uses credits per 1M tokens. OmniRoute stores the dollar-equivalent
+    // values below at the documented conversion of 25 credits per USD.
+    "gpt-5.6-sol": GPT_5_6_SOL_PRICING,
+    "gpt-5.6-sol-ultra": GPT_5_6_SOL_PRICING,
+    "gpt-5.6-sol-max": GPT_5_6_SOL_PRICING,
+    "gpt-5.6-sol-xhigh": GPT_5_6_SOL_PRICING,
+    "gpt-5.6-sol-high": GPT_5_6_SOL_PRICING,
+    "gpt-5.6-sol-medium": GPT_5_6_SOL_PRICING,
+    "gpt-5.6-sol-low": GPT_5_6_SOL_PRICING,
+    "gpt-5.6-terra": GPT_5_6_TERRA_PRICING,
+    "gpt-5.6-terra-ultra": GPT_5_6_TERRA_PRICING,
+    "gpt-5.6-terra-max": GPT_5_6_TERRA_PRICING,
+    "gpt-5.6-terra-xhigh": GPT_5_6_TERRA_PRICING,
+    "gpt-5.6-terra-high": GPT_5_6_TERRA_PRICING,
+    "gpt-5.6-terra-medium": GPT_5_6_TERRA_PRICING,
+    "gpt-5.6-terra-low": GPT_5_6_TERRA_PRICING,
+    "gpt-5.6-luna": GPT_5_6_LUNA_PRICING,
+    "gpt-5.6-luna-max": GPT_5_6_LUNA_PRICING,
+    "gpt-5.6-luna-xhigh": GPT_5_6_LUNA_PRICING,
+    "gpt-5.6-luna-high": GPT_5_6_LUNA_PRICING,
+    "gpt-5.6-luna-medium": GPT_5_6_LUNA_PRICING,
+    "gpt-5.6-luna-low": GPT_5_6_LUNA_PRICING,
     // GPT 5.5
     "gpt-5.5": GPT_5_5_PRICING,
     "gpt5.5": GPT_5_5_PRICING,
@@ -80,42 +108,6 @@ export const DEFAULT_PRICING_OAUTH = {
     "gpt-5.5-medium": GPT_5_5_PRICING,
     "gpt-5.5-low": GPT_5_5_PRICING,
     "gpt-5.5-none": GPT_5_5_PRICING,
-    // GPT 5.4
-    "gpt-5.4": {
-      input: 5.0,
-      output: 20.0,
-      cached: 2.5,
-      reasoning: 30.0,
-      cache_creation: 5.0,
-    },
-    "gpt5.4": {
-      input: 5.0,
-      output: 20.0,
-      cached: 2.5,
-      reasoning: 30.0,
-      cache_creation: 5.0,
-    },
-    // T12: fallback pricing for gpt-5.4 mini variants
-    "gpt-5.4-mini": {
-      input: 1.5,
-      output: 6.0,
-      cached: 0.75,
-      reasoning: 9.0,
-      cache_creation: 1.5,
-    },
-    "gpt5.4-mini": {
-      input: 1.5,
-      output: 6.0,
-      cached: 0.75,
-      reasoning: 9.0,
-      cache_creation: 1.5,
-    },
-    // gpt-5.4 reasoning-effort variants share the gpt-5.4 tier (registry exposes
-    // -xhigh/-high/-medium/-low; without these rows they resolved to $0).
-    "gpt-5.4-xhigh": GPT_5_3_CODEX_PRICING,
-    "gpt-5.4-high": GPT_5_3_CODEX_PRICING,
-    "gpt-5.4-medium": GPT_5_3_CODEX_PRICING,
-    "gpt-5.4-low": GPT_5_3_CODEX_PRICING,
     // GPT 5.3 Codex family (all same pricing tier)
     "gpt-5.3-codex-spark": GPT_5_3_CODEX_PRICING,
     "gpt-5.3-codex": GPT_5_3_CODEX_PRICING,

src/shared/constants/pricing/shared-tiers.ts

@@ -17,6 +17,30 @@ export const GPT_5_5_PRICING = {
   cache_creation: 5.0,
 };
 
+export const GPT_5_6_SOL_PRICING = {
+  input: 5.0,
+  output: 30.0,
+  cached: 0.5,
+  reasoning: 30.0,
+  cache_creation: 6.25,
+};
+
+export const GPT_5_6_TERRA_PRICING = {
+  input: 2.5,
+  output: 15.0,
+  cached: 0.25,
+  reasoning: 15.0,
+  cache_creation: 3.125,
+};
+
+export const GPT_5_6_LUNA_PRICING = {
+  input: 1.0,
+  output: 6.0,
+  cached: 0.1,
+  reasoning: 6.0,
+  cache_creation: 1.25,
+};
+
 export const CLAUDE_FABLE_5_PRICING = {
   input: 15.0,
   output: 75.0,

src/shared/reasoning/effortStandardization.ts

@@ -11,16 +11,38 @@ import { z } from "zod";
  * provider-agnostic pair of request fields and folds them onto the fields the existing
  * mappers already read.
  *
- * Canonical effort vocabulary — the SAME five values used everywhere else in the codebase
- * (`providerSpecificData.ts` CODEX_REASONING_EFFORT_VALUES, `vscode/reasoningMetadata.ts`
- * KNOWN_REASONING_EFFORTS, `modelSpecs.ts`). We deliberately REUSE this set instead of
- * inventing a parallel Low/Medium/High/Extra/Max enum that would diverge from the rest of
- * the codebase.
+ * The provider-agnostic vocabulary remains five values. Provider-native additions such as
+ * Codex GPT-5.6 Max and Ultra are exposed separately without widening this request contract.
  */
 export const CANONICAL_EFFORT_VALUES = ["none", "low", "medium", "high", "xhigh"] as const;
 
 export type CanonicalEffort = (typeof CANONICAL_EFFORT_VALUES)[number];
 
+/** Add provider-native GPT-5.6 effort levels without widening the global request vocabulary. */
+export function extendCodexGpt56EffortValues(
+  provider: string | null | undefined,
+  model: string | null | undefined,
+  baseValues: readonly string[]
+): string[] {
+  const values = [...baseValues];
+  const normalizedProvider = provider?.trim().toLowerCase();
+  const normalizedModel = model
+    ?.trim()
+    .toLowerCase()
+    .replace(/^(?:codex|cx)\//, "");
+  if (!normalizedModel || (normalizedProvider !== "codex" && normalizedProvider !== "cx")) {
+    return values;
+  }
+
+  const match = normalizedModel.match(
+    /^gpt-5\.6-(sol|terra|luna)(?:-(?:none|low|medium|high|xhigh|max|ultra))?$/
+  );
+  if (!match) return values;
+
+  const additions = match[1] === "luna" ? ["max"] : ["max", "ultra"];
+  return [...new Set([...values, ...additions])];
+}
+
 /**
  * UI-facing tier synonyms mapped onto the canonical set. The issue (#6241) requested a
  * 5-tier UI vocabulary (Low / Medium / High / Extra / Max); that request collapses onto
@@ -95,18 +117,13 @@ export function normalizeReasoningRequest<T>(body: T): T {
 
   const reasoning = body.reasoning;
   const clientSetReasoningEffort = body.reasoning_effort !== undefined;
-  const clientSetReasoningObjEffort =
-    isPlainObject(reasoning) && reasoning.effort !== undefined;
+  const clientSetReasoningObjEffort = isPlainObject(reasoning) && reasoning.effort !== undefined;
 
   const next: Record<string, unknown> = { ...body };
 
   // Canonical effort → the fields the mappers read. Skip entirely if the client already
   // expressed a reasoning effort (either shape) so client intent is preserved.
-  if (
-    canonicalEffort !== undefined &&
-    !clientSetReasoningEffort &&
-    !clientSetReasoningObjEffort
-  ) {
+  if (canonicalEffort !== undefined && !clientSetReasoningEffort && !clientSetReasoningObjEffort) {
     next.reasoning_effort = canonicalEffort;
     next.reasoning = {
       ...(isPlainObject(reasoning) ? reasoning : {}),

src/shared/services/codexCatalogRevalidation.ts

@@ -0,0 +1,395 @@
+/**
+ * Codex catalog revalidation (Codex provider only)
+ *
+ * Runs scrub + live re-sync only in three cases:
+ *  1) first-start — no version marker yet
+ *  2) upgrade — app version marker changed after update/reboot
+ *  3) init — setup/onboarding just completed (explicit trigger)
+ *
+ * Success log (single line): kill deprecated models complete.
+ */
+
+import { readFileSync } from "node:fs";
+import path from "node:path";
+import { isCodexDiscoveryModelExcluded } from "@/shared/services/codexDiscoveryPolicy";
+import {
+  getSyncedAvailableModelsForConnection,
+  replaceSyncedAvailableModelsForConnection,
+  type SyncedAvailableModel,
+} from "@/lib/db/models";
+import { getProviderConnections } from "@/lib/db/providers";
+import { getSettings, updateSettings } from "@/lib/db/settings";
+
+export const CODEX_CATALOG_REVALIDATED_VERSION_KEY = "codex_catalog_revalidated_version";
+
+export type CodexCatalogRevalidationReason = "first-start" | "upgrade" | "init";
+
+type AppVersionOptions = {
+  runtimeRoot?: string;
+  packageVersion?: string | null;
+};
+
+function readNonEmptyTextFile(filePath: string): string | null {
+  try {
+    const value = readFileSync(filePath, "utf8").trim();
+    return value || null;
+  } catch {
+    return null;
+  }
+}
+
+function readInstalledPackageVersion(runtimeRoot: string): string | null {
+  try {
+    const pkg = JSON.parse(readFileSync(path.join(runtimeRoot, "package.json"), "utf8")) as {
+      version?: unknown;
+    };
+    return typeof pkg.version === "string" && pkg.version.trim() ? pkg.version.trim() : null;
+  } catch {
+    return null;
+  }
+}
+
+/** Resolve a stable, source-qualified app identity for upgrade detection. */
+export function resolveCodexCatalogAppVersion(
+  env: NodeJS.ProcessEnv = process.env,
+  options: AppVersionOptions = {}
+): string | null {
+  for (const value of [env.OMNIROUTE_BUILD_SHA, env.BUILD_SHA]) {
+    if (typeof value === "string" && value.trim()) return `build:${value.trim()}`;
+  }
+
+  const runtimeRoot = options.runtimeRoot || process.cwd();
+  const buildSha = readNonEmptyTextFile(path.join(runtimeRoot, "BUILD_SHA"));
+  if (buildSha) return `build:${buildSha}`;
+
+  for (const buildIdPath of [
+    path.join(runtimeRoot, ".build", "next", "BUILD_ID"),
+    path.join(runtimeRoot, ".next", "BUILD_ID"),
+  ]) {
+    const buildId = readNonEmptyTextFile(buildIdPath);
+    if (buildId) return `next:${buildId}`;
+  }
+
+  const envPackageVersion = env.npm_package_version || env.OMNIROUTE_VERSION;
+  if (typeof envPackageVersion === "string" && envPackageVersion.trim()) {
+    return `pkg:${envPackageVersion.trim()}`;
+  }
+
+  const hasPackageVersionOverride = Object.prototype.hasOwnProperty.call(options, "packageVersion");
+  const packageVersion = hasPackageVersionOverride
+    ? options.packageVersion
+    : readInstalledPackageVersion(runtimeRoot);
+  return typeof packageVersion === "string" && packageVersion.trim()
+    ? `pkg:${packageVersion.trim()}`
+    : null;
+}
+
+/**
+ * Pure: map stored marker + current version → boot trigger, or null to skip.
+ * - no marker → first-start
+ * - marker !== version → upgrade
+ * - else → null (do nothing on this boot)
+ */
+export function resolveBootRevalidationReason(
+  previousVersion: string | null | undefined,
+  appVersion: string
+): CodexCatalogRevalidationReason | null {
+  if (!previousVersion || !String(previousVersion).trim()) return "first-start";
+  if (String(previousVersion).trim() !== appVersion) return "upgrade";
+  return null;
+}
+
+/** Pure helper: drop denylisted ids from a synced model list. */
+export function scrubSyncedModelsWithCodexDenylist(models: SyncedAvailableModel[]): {
+  kept: SyncedAvailableModel[];
+  removedIds: string[];
+} {
+  const removedIds: string[] = [];
+  const kept: SyncedAvailableModel[] = [];
+  for (const model of models) {
+    if (!model?.id) continue;
+    if (isCodexDiscoveryModelExcluded({ id: model.id })) {
+      removedIds.push(model.id);
+      continue;
+    }
+    kept.push(model);
+  }
+  return { kept, removedIds };
+}
+
+export type CodexCatalogScrubResult = {
+  connections: number;
+  connectionsChanged: number;
+  modelsRemoved: number;
+  removedIds: string[];
+};
+
+/** Offline pass: rewrite persisted Codex synced catalogs through the denylist. */
+export async function scrubCodexPersistedCatalogs(): Promise<CodexCatalogScrubResult> {
+  const connections = await getProviderConnections({ provider: "codex" });
+  const removedIds = new Set<string>();
+  let connectionsChanged = 0;
+
+  for (const connection of connections) {
+    const connectionId = String(connection.id || "");
+    if (!connectionId) continue;
+    const existing = await getSyncedAvailableModelsForConnection("codex", connectionId);
+    if (existing.length === 0) continue;
+    const { kept, removedIds: removed } = scrubSyncedModelsWithCodexDenylist(existing);
+    if (removed.length === 0) continue;
+    await replaceSyncedAvailableModelsForConnection("codex", connectionId, kept);
+    connectionsChanged += 1;
+    for (const id of removed) removedIds.add(id);
+  }
+
+  return {
+    connections: connections.length,
+    connectionsChanged,
+    modelsRemoved: removedIds.size,
+    removedIds: Array.from(removedIds).sort(),
+  };
+}
+
+async function listActiveCodexConnectionIds(): Promise<Array<{ id: string; name?: string }>> {
+  const connections = await getProviderConnections({ provider: "codex" });
+  return connections
+    .filter((conn) => conn.isActive !== false)
+    .map((conn) => ({
+      id: String(conn.id),
+      name: typeof conn.name === "string" ? conn.name : undefined,
+    }))
+    .filter((conn) => conn.id.length > 0);
+}
+
+export async function waitForLoopbackHttpReady(options?: {
+  apiBaseUrl?: string;
+  maxWaitMs?: number;
+  pollMs?: number;
+}): Promise<void> {
+  const maxWaitMs = options?.maxWaitMs ?? 15_000;
+  const pollMs = options?.pollMs ?? 50;
+  const { fetchModelSyncInternal, resolveModelSyncInternalBaseUrl } =
+    await import("./modelSyncScheduler");
+  const baseUrl = resolveModelSyncInternalBaseUrl(options?.apiBaseUrl);
+  const deadline = Date.now() + maxWaitMs;
+  let lastErr: unknown;
+
+  while (Date.now() < deadline) {
+    try {
+      const res = await fetchModelSyncInternal(
+        `${baseUrl}/api/providers/__readiness_probe__/models`,
+        {
+          redirect: "error",
+          signal: AbortSignal.timeout(1_500),
+        }
+      );
+      if (res.status >= 200 && res.status < 600) return;
+    } catch (err) {
+      lastErr = err;
+    }
+    await new Promise((r) => setTimeout(r, pollMs));
+  }
+
+  throw new Error(
+    `loopback not ready within ${maxWaitMs}ms: ${
+      lastErr instanceof Error ? lastErr.message : String(lastErr)
+    }`
+  );
+}
+
+export async function liveResyncCodexConnections(
+  apiBaseUrl?: string
+): Promise<{ attempted: number; succeeded: number }> {
+  const connections = await listActiveCodexConnectionIds();
+  if (connections.length === 0) {
+    return { attempted: 0, succeeded: 0 };
+  }
+
+  const { buildModelSyncInternalHeaders, fetchModelSyncInternal, resolveModelSyncInternalBaseUrl } =
+    await import("./modelSyncScheduler");
+  const base = resolveModelSyncInternalBaseUrl(apiBaseUrl);
+  const results = await Promise.allSettled(
+    connections.map(async (conn) => {
+      const res = await fetchModelSyncInternal(
+        `${base}/api/providers/${conn.id}/sync-models?quiet=1`,
+        {
+          method: "POST",
+          redirect: "error",
+          headers: {
+            "Content-Type": "application/json",
+            ...buildModelSyncInternalHeaders(),
+          },
+        }
+      );
+      if (!res.ok) {
+        throw new Error(`HTTP ${res.status}`);
+      }
+      return true;
+    })
+  );
+
+  const succeeded = results.filter((r) => r.status === "fulfilled").length;
+  return { attempted: connections.length, succeeded };
+}
+
+async function readPreviousVersionMarker(): Promise<string | null> {
+  try {
+    const settings = await getSettings();
+    const raw = settings?.[CODEX_CATALOG_REVALIDATED_VERSION_KEY];
+    return typeof raw === "string" && raw.trim() ? raw.trim() : null;
+  } catch {
+    return null;
+  }
+}
+
+async function writeVersionMarker(appVersion: string): Promise<boolean> {
+  try {
+    await updateSettings({ [CODEX_CATALOG_REVALIDATED_VERSION_KEY]: appVersion });
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+export type CodexCatalogRevalidationOutcome = {
+  complete: boolean;
+  attempted: number;
+  succeeded: number;
+};
+
+export async function executeCodexCatalogRevalidation(options: {
+  appVersion: string | null;
+  scrub: () => Promise<unknown>;
+  waitForReady: () => Promise<void>;
+  liveResync: () => Promise<{ attempted: number; succeeded: number }>;
+  writeMarker: (appVersion: string) => Promise<boolean>;
+  logSuccess: () => void;
+}): Promise<CodexCatalogRevalidationOutcome> {
+  await options.scrub();
+
+  try {
+    await options.waitForReady();
+  } catch {
+    return { complete: false, attempted: 0, succeeded: 0 };
+  }
+
+  const syncResult = await options.liveResync();
+  if (syncResult.succeeded !== syncResult.attempted) {
+    return { complete: false, ...syncResult };
+  }
+
+  if (!options.appVersion) return { complete: false, ...syncResult };
+  const markerWritten = await options.writeMarker(options.appVersion);
+  if (!markerWritten) return { complete: false, ...syncResult };
+
+  options.logSuccess();
+  return { complete: true, ...syncResult };
+}
+
+type CodexCatalogRevalidationRequest = {
+  apiBaseUrl?: string;
+  reason: CodexCatalogRevalidationReason;
+};
+
+export function createCodexCatalogRevalidationCoordinator(
+  run: (options: CodexCatalogRevalidationRequest) => Promise<void>
+): (options: CodexCatalogRevalidationRequest) => Promise<void> {
+  let activeRun: Promise<void> | null = null;
+  let activeReason: CodexCatalogRevalidationReason | null = null;
+  let queuedInit: CodexCatalogRevalidationRequest | null = null;
+
+  return (options) => {
+    if (activeRun) {
+      if (options.reason === "init" && activeReason !== "init") {
+        queuedInit = options;
+      }
+      return activeRun;
+    }
+
+    activeRun = (async () => {
+      let current: CodexCatalogRevalidationRequest | null = options;
+      let firstError: unknown;
+
+      try {
+        while (current) {
+          activeReason = current.reason;
+          try {
+            await run(current);
+          } catch (error) {
+            firstError ??= error;
+          }
+          current = queuedInit;
+          queuedInit = null;
+        }
+      } finally {
+        activeRun = null;
+        activeReason = null;
+      }
+
+      if (firstError) throw firstError;
+    })();
+
+    return activeRun;
+  };
+}
+
+/**
+ * Run scrub + live re-sync for an explicit reason, then mark version.
+ * Operator-facing success log is a single line.
+ */
+async function performCodexCatalogRevalidation(
+  options: CodexCatalogRevalidationRequest
+): Promise<void> {
+  const appVersion = resolveCodexCatalogAppVersion();
+  const { resolveModelSyncInternalBaseUrl } = await import("./modelSyncScheduler");
+  const apiBaseUrl = resolveModelSyncInternalBaseUrl(options.apiBaseUrl);
+
+  await executeCodexCatalogRevalidation({
+    appVersion,
+    scrub: scrubCodexPersistedCatalogs,
+    waitForReady: () => waitForLoopbackHttpReady({ apiBaseUrl }),
+    liveResync: () => liveResyncCodexConnections(apiBaseUrl),
+    writeMarker: writeVersionMarker,
+    logSuccess: () => console.log("kill deprecated models complete."),
+  });
+}
+
+const requestCodexCatalogRevalidation = createCodexCatalogRevalidationCoordinator(
+  performCodexCatalogRevalidation
+);
+
+export function revalidateCodexCatalogs(options: CodexCatalogRevalidationRequest): Promise<void> {
+  return requestCodexCatalogRevalidation(options);
+}
+
+/** Boot path: only first-start or upgrade. */
+export async function revalidateCodexCatalogsOnStartup(options?: {
+  apiBaseUrl?: string;
+}): Promise<void> {
+  const appVersion = resolveCodexCatalogAppVersion();
+  const previousVersion = await readPreviousVersionMarker();
+  const reason = appVersion
+    ? resolveBootRevalidationReason(previousVersion, appVersion)
+    : "first-start";
+  if (!reason) return;
+  await revalidateCodexCatalogs({ apiBaseUrl: options?.apiBaseUrl, reason });
+}
+
+function scheduleRun(run: () => Promise<void>): void {
+  const timer = setTimeout(() => {
+    void run().catch(() => {
+      // silent — success line only on full success
+    });
+  }, 0);
+  timer.unref?.();
+}
+
+/** Fire-and-forget boot schedule (first-start / upgrade only). */
+export function scheduleCodexCatalogRevalidation(options?: { apiBaseUrl?: string }): void {
+  scheduleRun(() => revalidateCodexCatalogsOnStartup({ apiBaseUrl: options?.apiBaseUrl }));
+}
+
+/** Fire-and-forget after setup/onboarding completes. */
+export function scheduleCodexCatalogRevalidationAfterInit(options?: { apiBaseUrl?: string }): void {
+  scheduleRun(() => revalidateCodexCatalogs({ apiBaseUrl: options?.apiBaseUrl, reason: "init" }));
+}

src/shared/services/codexDiscoveryPolicy.ts

@@ -0,0 +1,30 @@
+/** Exact Codex model ids retired after discovery merge. */
+export const CODEX_DISCOVERY_EXCLUDED_IDS: ReadonlySet<string> = new Set([
+  // Reserved for one-off retired ids that do not share a clean prefix family.
+]);
+
+/**
+ * Codex model-id families retired after discovery merge. Delimiter-aware
+ * matching prevents prefixes such as `gpt-5.40` from being removed.
+ */
+export const CODEX_DISCOVERY_EXCLUDED_ID_PREFIXES: readonly string[] = ["gpt-5.4"];
+
+export type CodexDiscoveryModelIdentity = {
+  id?: unknown;
+};
+
+export function isCodexDiscoveryModelExcluded(model: CodexDiscoveryModelIdentity): boolean {
+  const id = typeof model?.id === "string" ? model.id.trim().toLowerCase() : "";
+  if (!id) return true;
+  if (CODEX_DISCOVERY_EXCLUDED_IDS.has(id)) return true;
+
+  return CODEX_DISCOVERY_EXCLUDED_ID_PREFIXES.some((prefix) => {
+    const normalizedPrefix = prefix.toLowerCase();
+    return (
+      id === normalizedPrefix ||
+      id.startsWith(`${normalizedPrefix}-`) ||
+      id.startsWith(`${normalizedPrefix}_`) ||
+      id.startsWith(`${normalizedPrefix}.`)
+    );
+  });
+}

src/shared/services/modelSyncScheduler.ts

@@ -9,6 +9,7 @@
  */
 
 import { randomUUID } from "node:crypto";
+import { Agent, buildConnector, fetch as undiciFetch, type Dispatcher } from "undici";
 import { getSettings, updateSettings } from "@/lib/localDb";
 import { getRuntimePorts } from "@/lib/runtime/ports";
 
@@ -16,13 +17,15 @@ const DEFAULT_INTERVAL_MS = 24 * 60 * 60 * 1000; // 24 hours
 const MODEL_SYNC_SETTING_KEY = "model_sync_last_run";
 const MODEL_SYNC_INTERNAL_AUTH_HEADER = "x-model-sync-internal-auth";
 
-const { dashboardPort } = getRuntimePorts();
+function normalizeInternalBasePath(value: string | undefined): string {
+  const trimmed = value?.trim();
+  if (!trimmed || trimmed === "/") return "";
+  if (!trimmed.startsWith("/") || /[?#\\]/.test(trimmed)) return "";
 
-const INTERNAL_BASE_URL =
-  process.env.BASE_URL ||
-  process.env.NEXT_PUBLIC_BASE_URL ||
-  process.env.NEXT_PUBLIC_APP_URL ||
-  `http://127.0.0.1:${dashboardPort}`;
+  const segments = trimmed.split("/").filter(Boolean);
+  if (segments.some((segment) => segment === "." || segment === "..")) return "";
+  return `/${segments.join("/")}`;
+}
 
 /**
  * Trusted origin for server-internal self-fetches (model sync, auto-discovery).
@@ -34,9 +37,80 @@ const INTERNAL_BASE_URL =
  * this loopback/env-pinned origin instead.
  */
 export function getModelSyncInternalBaseUrl(): string {
-  return INTERNAL_BASE_URL;
+  return resolveModelSyncInternalBaseUrl();
 }
 
+export function resolveModelSyncInternalBaseUrl(_candidate?: string): string {
+  const { dashboardPort } = getRuntimePorts();
+  const nativeTls = process.env.OMNIROUTE_INTERNAL_SCHEME === "https";
+  const origin = nativeTls
+    ? `https://localhost:${dashboardPort}`
+    : `http://127.0.0.1:${dashboardPort}`;
+  return `${origin}${normalizeInternalBasePath(process.env.OMNIROUTE_BASE_PATH)}`;
+}
+
+export function createPinnedModelSyncTlsConnector(
+  connect: buildConnector.connector = buildConnector({ servername: "localhost" })
+): buildConnector.connector {
+  return (options, callback) =>
+    connect(
+      {
+        ...options,
+        host: "localhost",
+        hostname: "127.0.0.1",
+        servername: "localhost",
+      },
+      callback
+    );
+}
+
+let pinnedModelSyncTlsDispatcher: Dispatcher | null = null;
+
+function getPinnedModelSyncTlsDispatcher(): Dispatcher {
+  if (!pinnedModelSyncTlsDispatcher) {
+    pinnedModelSyncTlsDispatcher = new Agent({
+      connect: createPinnedModelSyncTlsConnector(),
+      connections: 8,
+      pipelining: 0,
+    });
+  }
+  return pinnedModelSyncTlsDispatcher;
+}
+
+const fetchWithDispatcher = undiciFetch as unknown as (
+  input: RequestInfo | URL,
+  init: RequestInit & { dispatcher: Dispatcher }
+) => Promise<Response>;
+
+export const fetchModelSyncInternal: typeof fetch = async (input, init = {}) => {
+  const inputUrl =
+    typeof input === "string" || input instanceof URL ? new URL(input) : new URL(input.url);
+  const expectedBase = new URL(getModelSyncInternalBaseUrl());
+  if (
+    inputUrl.protocol !== expectedBase.protocol ||
+    inputUrl.hostname !== expectedBase.hostname ||
+    inputUrl.port !== expectedBase.port ||
+    inputUrl.username ||
+    inputUrl.password
+  ) {
+    throw new TypeError("model sync internal fetch must target the active dashboard listener");
+  }
+
+  const basePath = expectedBase.pathname === "/" ? "" : expectedBase.pathname;
+  if (basePath && inputUrl.pathname !== basePath && !inputUrl.pathname.startsWith(`${basePath}/`)) {
+    throw new TypeError("model sync internal fetch must stay under the configured base path");
+  }
+
+  const requestInit = { ...init, redirect: "error" as const };
+  if (inputUrl.protocol === "https:") {
+    return fetchWithDispatcher(inputUrl, {
+      ...requestInit,
+      dispatcher: getPinnedModelSyncTlsDispatcher(),
+    });
+  }
+  return globalThis.fetch(inputUrl.href, requestInit);
+};
+
 const globalState = globalThis as typeof globalThis & {
   __omnirouteModelSyncInternalAuthToken?: string;
 };
@@ -78,14 +152,22 @@ async function getAutoSyncConnections(): Promise<
   try {
     const { getProviderConnections } = await import("@/lib/localDb");
     const connections = await getProviderConnections();
-    return connections.filter((conn: any) => {
-      if (!conn.isActive && conn.isActive !== undefined) return false;
+    const autoSyncConnections: Array<{ id: string; provider: string; name?: string }> = [];
+    for (const conn of connections) {
+      if (!conn.isActive && conn.isActive !== undefined) continue;
       const psd =
         conn.providerSpecificData && typeof conn.providerSpecificData === "object"
-          ? conn.providerSpecificData
+          ? (conn.providerSpecificData as Record<string, unknown>)
           : {};
-      return psd.autoSync === true;
-    });
+      if (psd.autoSync !== true) continue;
+      if (typeof conn.id !== "string" || typeof conn.provider !== "string") continue;
+      autoSyncConnections.push({
+        id: conn.id,
+        provider: conn.provider,
+        ...(typeof conn.name === "string" ? { name: conn.name } : {}),
+      });
+    }
+    return autoSyncConnections;
   } catch (err) {
     console.warn("[ModelSync] Failed to load connections:", (err as Error).message);
     return [];
@@ -101,13 +183,17 @@ async function syncConnectionModels(
   baseUrl: string
 ): Promise<boolean> {
   try {
-    const res = await fetch(`${baseUrl}/api/providers/${connectionId}/sync-models`, {
-      method: "POST",
-      headers: {
-        "Content-Type": "application/json",
-        ...buildModelSyncInternalHeaders(),
-      },
-    });
+    const res = await fetchModelSyncInternal(
+      `${baseUrl}/api/providers/${connectionId}/sync-models`,
+      {
+        method: "POST",
+        redirect: "error",
+        headers: {
+          "Content-Type": "application/json",
+          ...buildModelSyncInternalHeaders(),
+        },
+      }
+    );
     if (!res.ok) {
       console.warn(
         `[ModelSync] ${provider} (${connectionId.slice(0, 8)}): sync returned ${res.status}`
@@ -177,7 +263,7 @@ async function runSyncCycle(apiBaseUrl: string): Promise<void> {
  * @param intervalMs — sync interval in milliseconds (default: 24h)
  */
 export function startModelSyncScheduler(
-  apiBaseUrl = INTERNAL_BASE_URL,
+  apiBaseUrl = getModelSyncInternalBaseUrl(),
   intervalMs = DEFAULT_INTERVAL_MS
 ): void {
   if (schedulerTimer) {
@@ -189,15 +275,25 @@ export function startModelSyncScheduler(
   const envHours = parseInt(process.env.MODEL_SYNC_INTERVAL_HOURS ?? "", 10);
   const effectiveIntervalMs =
     !isNaN(envHours) && envHours > 0 ? envHours * 60 * 60 * 1000 : intervalMs;
+  const trustedApiBaseUrl = resolveModelSyncInternalBaseUrl(apiBaseUrl);
 
   console.log(`[ModelSync] Scheduler started — interval: ${effectiveIntervalMs / 3_600_000}h`);
 
   // Run immediately on startup (staggered by 5s to avoid startup congestion)
-  const startupDelay = setTimeout(() => runSyncCycle(apiBaseUrl), 5_000);
+  const startupDelay = setTimeout(() => runSyncCycle(trustedApiBaseUrl), 5_000);
   startupDelay.unref?.();
 
+  // Codex-only: revalidate catalog only on first-start or app upgrade (not every boot).
+  void import("./codexCatalogRevalidation")
+    .then(({ scheduleCodexCatalogRevalidation }) => {
+      scheduleCodexCatalogRevalidation({ apiBaseUrl: trustedApiBaseUrl });
+    })
+    .catch(() => {
+      // silent
+    });
+
   // Then run on the regular interval
-  schedulerTimer = setInterval(() => runSyncCycle(apiBaseUrl), effectiveIntervalMs);
+  schedulerTimer = setInterval(() => runSyncCycle(trustedApiBaseUrl), effectiveIntervalMs);
   schedulerTimer.unref?.();
 }
 

src/shared/validation/providerSpecificData.ts

@@ -13,7 +13,7 @@ function isHttpUrl(value: string): boolean {
   }
 }
 
-const CODEX_REASONING_EFFORT_VALUES = new Set(["none", "low", "medium", "high", "xhigh"]);
+const CODEX_REASONING_EFFORT_VALUES = new Set(["none", "low", "medium", "high", "xhigh", "max"]);
 const REQUEST_DEFAULT_SERVICE_TIER_VALUES = new Set(["default", "priority", "fast", "flex"]);
 
 export function validateProviderSpecificData(
@@ -131,7 +131,7 @@ export function validateProviderSpecificData(
         ctx.addIssue({
           code: z.ZodIssueCode.custom,
           message:
-            "providerSpecificData.requestDefaults.reasoningEffort must be one of none, low, medium, high, xhigh",
+            "providerSpecificData.requestDefaults.reasoningEffort must be one of none, low, medium, high, xhigh, max",
           path: ["requestDefaults", "reasoningEffort"],
         });
       }

src/shared/validation/schemas/cli.ts

@@ -68,7 +68,7 @@ export const cliModelConfigSchema = z.object({
   baseUrl: z.string().trim().min(1, "baseUrl and model are required"),
   apiKey: z.string().nullable().optional(),
   model: z.string().trim().min(1, "baseUrl and model are required"),
-  reasoningEffort: z.enum(["none", "low", "medium", "high", "xhigh"]).optional(),
+  reasoningEffort: z.enum(["none", "low", "medium", "high", "xhigh", "max", "ultra"]).optional(),
   wireApi: z.enum(["chat", "responses"]).optional(),
   modelMappings: z.record(z.string().trim().min(1), z.string().trim().min(1)).optional(),
 });

src/shared/validation/schemas/misc.ts

@@ -14,7 +14,6 @@ import {
 } from "@/shared/constants/upstreamHeaders";
 import { MAX_TIMER_TIMEOUT_MS } from "@/shared/utils/runtimeTimeouts";
 
-
 export function isHttpUrl(value: string): boolean {
   try {
     const parsed = new URL(value);
@@ -24,7 +23,14 @@ export function isHttpUrl(value: string): boolean {
   }
 }
 
-export const CODEX_REASONING_EFFORT_VALUES = new Set(["none", "low", "medium", "high", "xhigh"]);
+export const CODEX_REASONING_EFFORT_VALUES = new Set([
+  "none",
+  "low",
+  "medium",
+  "high",
+  "xhigh",
+  "max",
+]);
 
 export const REQUEST_DEFAULT_SERVICE_TIER_VALUES = new Set(["default", "priority", "fast", "flex"]);
 
@@ -200,4 +206,4 @@ export const confirmedAccountSchema = z.object({
   fingerprint: z.string().min(1).max(100),
 });
 
-export type ConfirmedAccount = z.infer<typeof confirmedAccountSchema>;
+export type ConfirmedAccount = z.infer<typeof confirmedAccountSchema>;

src/sse/handlers/chat.ts

@@ -34,6 +34,7 @@ import {
 } from "@omniroute/open-sse/config/constants.ts";
 import { getTargetFormat } from "@omniroute/open-sse/services/provider.ts";
 import {
+  getModelsByProviderId,
   getModelTargetFormat,
   PROVIDER_ID_TO_ALIAS,
 } from "@omniroute/open-sse/config/providerModels.ts";
@@ -311,11 +312,7 @@ export async function handleChat(
       }
     }
     if (b.max_tokens !== undefined) {
-      if (
-        typeof b.max_tokens !== "number" ||
-        !Number.isInteger(b.max_tokens) ||
-        b.max_tokens < 1
-      ) {
+      if (typeof b.max_tokens !== "number" || !Number.isInteger(b.max_tokens) || b.max_tokens < 1) {
         return badParam("max_tokens", "must be a positive integer");
       }
     }
@@ -402,9 +399,14 @@ export async function handleChat(
   // Image-only models live in IMAGE_PROVIDERS (open-sse/config/imageRegistry.ts)
   // and are served by /v1/images/generations. Forwarding them to a chat upstream
   // yielded confusing raw provider 400s (e.g. HuggingFace: "not a chat model").
-  // getImageModelEntry returns non-null only for models registered in the image
-  // registry — chat-only models (openai/gpt-4o, etc.) resolve to null and pass.
-  if (getImageModelEntry(modelStr)) {
+  // Models such as Codex GPT-5.5 support both chat and image generation, so an
+  // image-registry match is only image-only when the same provider/model pair is
+  // absent from the chat catalog.
+  const imageModel = getImageModelEntry(modelStr);
+  const isChatCatalogModel = imageModel
+    ? getModelsByProviderId(imageModel.provider).some((model) => model.id === imageModel.model)
+    : false;
+  if (imageModel && !isChatCatalogModel) {
     log.warn("CHAT", `Rejecting image-generation model on chat endpoint: ${modelStr}`);
     return errorResponse(
       HTTP_STATUS.BAD_REQUEST,

tests/integration/chat-pipeline.test.ts

@@ -698,7 +698,7 @@ test("chat pipeline applies Codex CLI fingerprint to OAuth responses requests",
   assert.equal(call.headers.Version, getCodexClientVersion());
   assert.equal(call.headers["Openai-Beta"], "responses=experimental");
   assert.equal(call.headers["X-Codex-Beta-Features"], "responses_websockets");
-  assert.equal(call.headers["User-Agent"], "codex-cli/0.144.0 (Windows 10.0.26200; x64)");
+  assert.equal(call.headers["User-Agent"], "codex-cli/0.144.1 (Windows 10.0.26200; x64)");
   assert.equal(call.headers["x-codex-window-id"], "conv_codex_fingerprint:0");
   assert.ok(call.headers["x-client-request-id"], "expected Codex request id header");
   assert.ok(call.headers["x-codex-turn-metadata"], "expected Codex turn metadata header");

tests/manual/image-generation.http

@@ -59,7 +59,7 @@ Authorization: Bearer {{OMNIROUTE_API_KEY}}
 Content-Type: application/json
 
 {
-  "model": "codex/gpt-5.4",
+  "model": "codex/gpt-5.6-sol",
   "prompt": "A happy red kitten, highly detailed fur, studio lighting",
   "n": 1,
   "size": "1024x1024",
@@ -90,4 +90,4 @@ Content-Type: application/json
   "stream": true
 }
 
-###
+###

tests/snapshots/provider/translate-path.json

@@ -28,7 +28,7 @@
       "apiKey": {
         "Accept": "text/event-stream",
         "Content-Type": "application/json",
-        "User-Agent": "claude-cli/2.1.195 (external, sdk-cli)",
+        "User-Agent": "claude-cli/2.1.207 (external, sdk-cli)",
         "X-Stainless-Arch": "<ARCH>",
         "X-Stainless-Lang": "js",
         "X-Stainless-OS": "MacOS",
@@ -47,7 +47,7 @@
       "nonStream": {
         "Accept": "application/json",
         "Content-Type": "application/json",
-        "User-Agent": "claude-cli/2.1.195 (external, sdk-cli)",
+        "User-Agent": "claude-cli/2.1.207 (external, sdk-cli)",
         "X-Stainless-Arch": "<ARCH>",
         "X-Stainless-Lang": "js",
         "X-Stainless-OS": "MacOS",
@@ -66,7 +66,7 @@
       "oauth": {
         "Accept": "text/event-stream",
         "Content-Type": "application/json",
-        "User-Agent": "claude-cli/2.1.195 (external, sdk-cli)",
+        "User-Agent": "claude-cli/2.1.207 (external, sdk-cli)",
         "X-Stainless-Arch": "<ARCH>",
         "X-Stainless-Lang": "js",
         "X-Stainless-OS": "MacOS",
@@ -713,7 +713,7 @@
         "Anthropic-Dangerous-Direct-Browser-Access": "true",
         "Anthropic-Version": "2023-06-01",
         "Content-Type": "application/json",
-        "User-Agent": "claude-cli/2.1.195 (external, cli)",
+        "User-Agent": "claude-cli/2.1.207 (external, cli)",
         "X-App": "cli",
         "X-Stainless-Arch": "<ARCH>",
         "X-Stainless-Helper-Method": "stream",
@@ -731,7 +731,7 @@
         "Anthropic-Dangerous-Direct-Browser-Access": "true",
         "Anthropic-Version": "2023-06-01",
         "Content-Type": "application/json",
-        "User-Agent": "claude-cli/2.1.195 (external, cli)",
+        "User-Agent": "claude-cli/2.1.207 (external, cli)",
         "X-App": "cli",
         "X-Stainless-Arch": "<ARCH>",
         "X-Stainless-Helper-Method": "stream",
@@ -750,7 +750,7 @@
         "Anthropic-Dangerous-Direct-Browser-Access": "true",
         "Anthropic-Version": "2023-06-01",
         "Content-Type": "application/json",
-        "User-Agent": "claude-cli/2.1.195 (external, cli)",
+        "User-Agent": "claude-cli/2.1.207 (external, cli)",
         "X-App": "cli",
         "X-Stainless-Arch": "<ARCH>",
         "X-Stainless-Helper-Method": "stream",
@@ -966,16 +966,16 @@
         "Authorization": "Bearer <TOK>",
         "Content-Type": "application/json",
         "Openai-Beta": "responses=experimental",
-        "User-Agent": "codex-cli/0.144.0 (<OS>; <ARCH>)",
-        "Version": "0.144.0",
+        "User-Agent": "codex-cli/0.144.1 (<OS>; <ARCH>)",
+        "Version": "0.144.1",
         "X-Codex-Beta-Features": "responses_websockets"
       },
       "nonStream": {
         "Authorization": "Bearer <TOK>",
         "Content-Type": "application/json",
         "Openai-Beta": "responses=experimental",
-        "User-Agent": "codex-cli/0.144.0 (<OS>; <ARCH>)",
-        "Version": "0.144.0",
+        "User-Agent": "codex-cli/0.144.1 (<OS>; <ARCH>)",
+        "Version": "0.144.1",
         "X-Codex-Beta-Features": "responses_websockets"
       },
       "oauth": {
@@ -983,8 +983,8 @@
         "Authorization": "Bearer <TOK>",
         "Content-Type": "application/json",
         "Openai-Beta": "responses=experimental",
-        "User-Agent": "codex-cli/0.144.0 (<OS>; <ARCH>)",
-        "Version": "0.144.0",
+        "User-Agent": "codex-cli/0.144.1 (<OS>; <ARCH>)",
+        "Version": "0.144.1",
         "X-Codex-Beta-Features": "responses_websockets"
       }
     },

tests/unit/chat-context-relay.test.ts

@@ -9,7 +9,7 @@ const { BaseExecutor, buildRequest, combosDb, handleChat, resetStorage, waitFor,
 const providersDb = await import("../../src/lib/db/providers.ts");
 const handoffDb = await import("../../src/lib/db/contextHandoffs.ts");
 
-function buildResponsesResponse(text = "ok", model = "gpt-5.4") {
+function buildResponsesResponse(text = "ok", model = "gpt-5.6-sol") {
   return new Response(
     JSON.stringify({
       id: "resp_context_relay",
@@ -112,7 +112,7 @@ test("handleChat generates and injects context-relay handoffs across Codex accou
       handoffThreshold: 0.85,
       maxMessagesForSummary: 12,
     },
-    models: ["codex/gpt-5.4"],
+    models: ["codex/gpt-5.6-sol"],
   });
 
   const upstreamBodies = [];
@@ -142,12 +142,12 @@ test("handleChat generates and injects context-relay handoffs across Codex accou
           taskProgress: "Runtime and UI are wired; tests are next",
           activeEntities: ["open-sse/services/combo.ts", "src/sse/handlers/chat.ts"],
         }),
-        "gpt-5.4"
+        "gpt-5.6-sol"
       );
     }
 
     upstreamBodies.push({ body, serializedBody });
-    return buildResponsesResponse("relay-success", "gpt-5.4");
+    return buildResponsesResponse("relay-success", "gpt-5.6-sol");
   };
 
   const firstResponse = await handleChat(
@@ -226,7 +226,7 @@ test("handleChat injects context-relay handoffs during live failover for Respons
       handoffThreshold: 0.85,
       maxMessagesForSummary: 12,
     },
-    models: ["codex/gpt-5.4"],
+    models: ["codex/gpt-5.6-sol"],
   });
 
   const upstreamBodies = [];
@@ -255,7 +255,7 @@ test("handleChat injects context-relay handoffs during live failover for Respons
           taskProgress: "Continue after the first account is exhausted",
           activeEntities: ["src/sse/handlers/chat.ts", "open-sse/services/contextHandoff.ts"],
         }),
-        "gpt-5.4"
+        "gpt-5.6-sol"
       );
     }
 
@@ -271,7 +271,7 @@ test("handleChat injects context-relay handoffs during live failover for Respons
       }
     }
 
-    return buildResponsesResponse("relay-success", "gpt-5.4");
+    return buildResponsesResponse("relay-success", "gpt-5.6-sol");
   };
 
   const firstResponse = await handleChat(
@@ -331,9 +331,7 @@ test("handleChat injects context-relay handoffs during live failover for Respons
   assert.equal(secondResponse.status, 200);
 
   const relayedSecondaryCall = upstreamBodies.find(
-    (call) =>
-      call.authHeader === "Bearer token-b" &&
-      typeof call.body.instructions === "string"
+    (call) => call.authHeader === "Bearer token-b" && typeof call.body.instructions === "string"
   );
 
   assert.ok(relayedSecondaryCall, "secondary account should receive a request after primary 429");

tests/unit/chat-rejects-image-only-model.test.ts

@@ -71,3 +71,23 @@ test("POST /v1/chat/completions with a chat model still reaches routing (guard i
     assert.doesNotMatch(msg, /image-generation model/i, "chat model must not trip the image guard");
   }
 });
+
+test("POST /v1/chat/completions allows a model registered for both chat and image generation", async () => {
+  const request = buildRequest({
+    body: {
+      model: "codex/gpt-5.6-sol",
+      messages: [{ role: "user", content: "hi" }],
+    },
+  });
+
+  const res = await handleChat(request);
+  if (res.status === 400) {
+    const body = (await res.json()) as { error?: { message?: string } };
+    const msg = body?.error?.message || JSON.stringify(body);
+    assert.doesNotMatch(
+      msg,
+      /image-generation model/i,
+      "a model present in the chat catalog must not trip the image-only guard"
+    );
+  }
+});

tests/unit/chatgpt-web-handoff-resume.test.ts

@@ -0,0 +1,260 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+
+import type { TlsFetchOptions } from "../../open-sse/services/chatgptTlsClient.ts";
+
+const { ChatGptWebExecutor, __resetChatGptWebCachesForTesting } =
+  await import("../../open-sse/executors/chatgpt-web.ts");
+const { __setTlsFetchOverrideForTesting } =
+  await import("../../open-sse/services/chatgptTlsClient.ts");
+
+function makeHeaders(values: Record<string, string> = {}): Headers {
+  const headers = new Headers();
+  for (const [name, value] of Object.entries(values)) headers.set(name, value);
+  return headers;
+}
+
+function sseText(events: unknown[]): string {
+  return `${events.map((event) => `data: ${JSON.stringify(event)}\r\n\r\n`).join("")}data: [DONE]\r\n\r\n`;
+}
+
+type ResumeRequest = {
+  body: { conversation_id?: string; offset?: number };
+  headers: Record<string, string>;
+};
+
+function installHandoffMock(
+  finalText: string,
+  options: { firstResumeStatus?: number } = {}
+): {
+  calls: { conversationDetail: number; resume: ResumeRequest[] };
+  restore: () => void;
+} {
+  const calls = {
+    conversationDetail: 0,
+    resume: [] as ResumeRequest[],
+  };
+
+  __setTlsFetchOverrideForTesting(async (url: string, request: TlsFetchOptions = {}) => {
+    const target = String(url);
+    const json = (body: unknown, status = 200) => ({
+      status,
+      headers: makeHeaders({ "Content-Type": "application/json" }),
+      text: JSON.stringify(body),
+      body: null,
+    });
+
+    if (
+      (target === "https://chatgpt.com/" || target === "https://chatgpt.com") &&
+      (request.method ?? "GET") === "GET"
+    ) {
+      return {
+        status: 200,
+        headers: makeHeaders({ "Content-Type": "text/html" }),
+        text: '<html data-build="prod-test"><script src="/_next/static/chunks/main.js"></script></html>',
+        body: null,
+      };
+    }
+
+    if (target.includes("/api/auth/session")) {
+      return json({
+        accessToken: "jwt-test",
+        expires: new Date(Date.now() + 3_600_000).toISOString(),
+        user: { id: "account-test" },
+      });
+    }
+
+    if (target.includes("/sentinel/chat-requirements")) {
+      return json({ token: "requirements-token", proofofwork: { required: false } });
+    }
+
+    if (target.endsWith("/backend-api/f/conversation/resume")) {
+      const body = JSON.parse(request.body ?? "{}") as ResumeRequest["body"];
+      calls.resume.push({ body, headers: request.headers ?? {} });
+      if (options.firstResumeStatus && calls.resume.length === 1) {
+        return {
+          status: options.firstResumeStatus,
+          headers: makeHeaders({ "Content-Type": "text/plain" }),
+          text: "not ready",
+          body: null,
+        };
+      }
+      return {
+        status: 200,
+        headers: makeHeaders({ "Content-Type": "text/event-stream" }),
+        text: sseText([
+          {
+            conversation_id: "conversation-handoff",
+            message: {
+              id: "assistant-final",
+              author: { role: "assistant" },
+              content: { content_type: "text", parts: [finalText] },
+              status: "in_progress",
+            },
+          },
+          {
+            conversation_id: "conversation-handoff",
+            message: {
+              id: "assistant-final",
+              author: { role: "assistant" },
+              content: { content_type: "text", parts: [finalText] },
+              status: "finished_successfully",
+              end_turn: true,
+            },
+          },
+        ]),
+        body: null,
+      };
+    }
+
+    if (target.endsWith("/backend-api/f/conversation")) {
+      return {
+        status: 200,
+        headers: makeHeaders({ "Content-Type": "text/event-stream" }),
+        text: sseText([
+          {
+            type: "resume_conversation_token",
+            token: "resume-token",
+            conversation_id: "conversation-handoff",
+          },
+          {
+            type: "stream_handoff",
+            conversation_id: "conversation-handoff",
+            turn_exchange_id: "turn-handoff",
+            options: [
+              { type: "resume_sse_endpoint", topic_id: "conversation-turn-handoff" },
+              { type: "subscribe_ws_topic", topic_id: "conversation-turn-handoff" },
+            ],
+          },
+        ]),
+        body: null,
+      };
+    }
+
+    if (/\/backend-api\/conversation\/[^/?#]+$/.test(target)) {
+      calls.conversationDetail++;
+      return json(
+        {
+          detail: {
+            message: "You do not have access to this temporary conversation.",
+            code: "conversation_not_found",
+          },
+        },
+        404
+      );
+    }
+
+    // Browser warmup requests are non-fatal, but returning 200 keeps test logs quiet.
+    if (
+      target.includes("/backend-api/me") ||
+      target.includes("/backend-api/conversations?") ||
+      target.includes("/backend-api/models?")
+    ) {
+      return json({});
+    }
+
+    return { status: 404, headers: makeHeaders(), text: "not mocked", body: null };
+  });
+
+  return {
+    calls,
+    restore() {
+      __setTlsFetchOverrideForTesting(null);
+    },
+  };
+}
+
+test("ChatGPT Web Pro models resume Temporary Chat handoffs through native SSE", async (t) => {
+  for (const model of ["gpt-5.6-pro", "gpt-5.5-pro", "gpt-5.5-pro-extended"]) {
+    await t.test(model, async () => {
+      __resetChatGptWebCachesForTesting();
+      const expected = `RESUMED_${model}`;
+      const mock = installHandoffMock(expected);
+      try {
+        const executor = new ChatGptWebExecutor();
+        const result = await executor.execute({
+          model,
+          body: { messages: [{ role: "user", content: "hard problem" }] },
+          stream: false,
+          credentials: { apiKey: `cookie-${model}` },
+          signal: AbortSignal.timeout(20_000),
+          log: null,
+        });
+
+        assert.equal(result.response.status, 200);
+        const response = await result.response.json();
+        assert.equal(response.choices[0].message.content, expected);
+        assert.equal(mock.calls.resume.length, 1);
+        assert.deepEqual(mock.calls.resume[0].body, {
+          conversation_id: "conversation-handoff",
+          offset: 0,
+        });
+        assert.equal(mock.calls.resume[0].headers["x-conduit-token"], "resume-token");
+        assert.equal(mock.calls.conversationDetail, 0);
+      } finally {
+        mock.restore();
+      }
+    });
+  }
+});
+
+test("ChatGPT Web handoff retries the next resume offset after a 404", async () => {
+  __resetChatGptWebCachesForTesting();
+  const mock = installHandoffMock("OFFSET_ONE_OK", { firstResumeStatus: 404 });
+  try {
+    const executor = new ChatGptWebExecutor();
+    const result = await executor.execute({
+      model: "gpt-5.6-pro",
+      body: { messages: [{ role: "user", content: "hard problem" }] },
+      stream: false,
+      credentials: { apiKey: "cookie-offset" },
+      signal: AbortSignal.timeout(20_000),
+      log: null,
+    });
+
+    assert.equal(result.response.status, 200);
+    const response = await result.response.json();
+    assert.equal(response.choices[0].message.content, "OFFSET_ONE_OK");
+    assert.deepEqual(
+      mock.calls.resume.map((call) => call.body.offset),
+      [0, 1]
+    );
+    assert.equal(mock.calls.conversationDetail, 0);
+  } finally {
+    mock.restore();
+  }
+});
+
+test("ChatGPT Web streaming appends the native resumed Pro answer", async () => {
+  __resetChatGptWebCachesForTesting();
+  const mock = installHandoffMock("STREAM_RESUME_OK");
+  try {
+    const executor = new ChatGptWebExecutor();
+    const result = await executor.execute({
+      model: "gpt-5.5-pro-extended",
+      body: { messages: [{ role: "user", content: "hard problem" }], stream: true },
+      stream: true,
+      credentials: { apiKey: "cookie-stream" },
+      signal: AbortSignal.timeout(20_000),
+      log: null,
+    });
+
+    assert.equal(result.response.status, 200);
+    const responseText = await result.response.text();
+    const content = responseText
+      .split("\n")
+      .filter((line) => line.startsWith("data: ") && line !== "data: [DONE]")
+      .map((line) => JSON.parse(line.slice(6)) as Record<string, unknown>)
+      .map((event) => {
+        const choices = event.choices as Array<{ delta?: { content?: string } }> | undefined;
+        return choices?.[0]?.delta?.content ?? "";
+      })
+      .join("");
+
+    assert.equal(content, "STREAM_RESUME_OK");
+    assert.equal(mock.calls.resume.length, 1);
+    assert.equal(mock.calls.conversationDetail, 0);
+  } finally {
+    mock.restore();
+  }
+});

tests/unit/chatgpt-web.test.ts

@@ -1245,21 +1245,26 @@ test("Provider registry: chatgpt-web exposes the current ChatGPT Web model catal
   assert.equal(entry.authHeader, "cookie");
 
   const ids = (entry.models || []).map((m) => m.id);
-  // Public OmniRoute ids stay in historical dot form even though ChatGPT's
-  // backend routes use dash-form slugs. Retired GPT-5/GPT-5.1 entries should
-  // stay out of this list.
+  // Retired GPT-5.4 and older entries stay out of the advertised catalog.
   assert.deepEqual(ids, [
-    "gpt-5.5-pro",
+    "gpt-5.6-pro",
+    "gpt-5.6-thinking",
     "gpt-5.5-pro-extended",
+    "gpt-5.5-pro",
     "gpt-5.5-thinking",
     "gpt-5.5",
-    "gpt-5.4-pro",
-    "gpt-5.4-thinking",
-    "gpt-5.4-thinking-mini",
-    "gpt-5.3",
-    "gpt-5.3-mini",
     "o3",
   ]);
+  assert.equal(
+    ids.some((id) => id.startsWith("gpt-5.4")),
+    false
+  );
+
+  const { MODEL_MAP } = await import("../../open-sse/executors/chatgpt-web/models.ts");
+  assert.equal(
+    Object.keys(MODEL_MAP).some((id) => id.startsWith("gpt-5.4") || id.startsWith("gpt-5-4")),
+    false
+  );
 });
 
 test("Executor MODEL_MAP: OmniRoute IDs translate to ChatGPT backend slugs", async () => {
@@ -1268,18 +1273,17 @@ test("Executor MODEL_MAP: OmniRoute IDs translate to ChatGPT backend slugs", asy
   try {
     const cases: Array<[string, string]> = [
       // Public catalog ids.
-      ["gpt-5.3", "gpt-5-3"],
+      ["gpt-5.6-pro", "gpt-5-6-pro"],
+      ["gpt-5.6-thinking", "gpt-5-6-thinking"],
       ["gpt-5.5-thinking", "gpt-5-5-thinking"],
-      ["gpt-5.4-thinking-mini", "gpt-5-4-t-mini"],
       ["gpt-5.5", "gpt-5-5"],
       ["gpt-5.5-pro", "gpt-5-5-pro"],
       ["gpt-5.5-pro-extended", "gpt-5-5-pro"],
-      ["gpt-5.4-pro", "gpt-5-4-pro"],
       ["o3", "o3"],
       // Backend dash-form slugs are still accepted for direct provider/model callers.
       ["gpt-5-3", "gpt-5-3"],
       ["gpt-5-5-thinking", "gpt-5-5-thinking"],
-      ["gpt-5-4-t-mini", "gpt-5-4-t-mini"],
+      ["gpt-5-6-pro", "gpt-5-6-pro"],
       ["gpt-5-5-pro", "gpt-5-5-pro"],
       ["gpt-5-5-pro-extended", "gpt-5-5-pro"],
     ];
@@ -1309,15 +1313,12 @@ test("MODEL_MAP drift guard: every advertised catalog id reaches ChatGPT as a ba
   const { getRegistryEntry } = await import("../../open-sse/config/providerRegistry.ts");
   const ids = (getRegistryEntry("chatgpt-web")?.models || []).map((m) => m.id);
   const expectedSlugById: Record<string, string> = {
-    "gpt-5.5-pro": "gpt-5-5-pro",
+    "gpt-5.6-pro": "gpt-5-6-pro",
+    "gpt-5.6-thinking": "gpt-5-6-thinking",
     "gpt-5.5-pro-extended": "gpt-5-5-pro",
+    "gpt-5.5-pro": "gpt-5-5-pro",
     "gpt-5.5-thinking": "gpt-5-5-thinking",
     "gpt-5.5": "gpt-5-5",
-    "gpt-5.4-pro": "gpt-5-4-pro",
-    "gpt-5.4-thinking": "gpt-5-4-thinking",
-    "gpt-5.4-thinking-mini": "gpt-5-4-t-mini",
-    "gpt-5.3": "gpt-5-3",
-    "gpt-5.3-mini": "gpt-5-3-mini",
     o3: "o3",
   };
   const m = installMockFetch();
@@ -1466,7 +1467,7 @@ test("thinking_effort: low/medium → PATCH with standard", async () => {
     try {
       const executor = new ChatGptWebExecutor();
       await executor.execute({
-        model: "gpt-5.4-thinking",
+        model: "gpt-5.6-thinking",
         body: { messages: [{ role: "user", content: "hi" }], reasoning_effort: effort },
         stream: false,
         credentials: { apiKey: `cookie-${effort}` },
@@ -1475,7 +1476,7 @@ test("thinking_effort: low/medium → PATCH with standard", async () => {
       });
       assert.equal(m.calls.userConfig, 1, `effort=${effort} should issue exactly one PATCH`);
       assert.match(m.calls.userConfigUrls[0], /thinking_effort=standard/, `${effort} → standard`);
-      assert.match(m.calls.userConfigUrls[0], /model_slug=gpt-5-4-thinking/);
+      assert.match(m.calls.userConfigUrls[0], /model_slug=gpt-5-6-thinking/);
     } finally {
       m.restore();
     }
@@ -1501,12 +1502,8 @@ test("thinking_effort: instant model never triggers PATCH even with reasoning_ef
   }
 });
 
-test("thinking_effort: bare chatgpt.com slug (e.g. gpt-5-4-t-mini) passed as model still PATCHes", async () => {
-  // Regression: the abbreviated dash-form slug "gpt-5-4-t-mini" doesn't
-  // carry the literal "thinking" substring, and isn't a key in MODEL_MAP
-  // (only its dot-form alias is), so a substring-only check would silently
-  // skip the PATCH for callers that send the chatgpt.com slug directly.
-  for (const bareSlug of ["gpt-5-4-t-mini", "gpt-5-5-thinking", "o3"]) {
+test("thinking_effort: bare chatgpt.com thinking slugs still PATCH", async () => {
+  for (const bareSlug of ["gpt-5-6-thinking", "gpt-5-5-thinking", "o3"]) {
     reset();
     const m = installMockFetch();
     try {
@@ -1559,7 +1556,7 @@ test("thinking_effort: providerSpecificData.thinkingEffort=extended overrides bo
   try {
     const executor = new ChatGptWebExecutor();
     await executor.execute({
-      model: "gpt-5.4-thinking-mini",
+      model: "gpt-5.6-thinking",
       body: {
         messages: [{ role: "user", content: "hi" }],
         reasoning_effort: "low", // would normally map to standard
@@ -1573,7 +1570,7 @@ test("thinking_effort: providerSpecificData.thinkingEffort=extended overrides bo
       log: null,
     });
     assert.equal(m.calls.userConfig, 1);
-    assert.match(m.calls.userConfigUrls[0], /model_slug=gpt-5-4-t-mini/);
+    assert.match(m.calls.userConfigUrls[0], /model_slug=gpt-5-6-thinking/);
     assert.match(m.calls.userConfigUrls[0], /thinking_effort=extended/);
   } finally {
     m.restore();
@@ -1676,12 +1673,12 @@ test("thinking_effort: PATCH failure is non-fatal — conversation request still
   }
 });
 
-test("Image registry: cgpt-web/gpt-5.3-instant routes to ChatGPT Web image handler", async () => {
+test("Image registry: cgpt-web/gpt-5.5 routes to ChatGPT Web image handler", async () => {
   const { parseImageModel, getImageProvider } =
     await import("../../open-sse/config/imageRegistry.ts");
-  const parsed = parseImageModel("cgpt-web/gpt-5.3-instant");
+  const parsed = parseImageModel("cgpt-web/gpt-5.5");
   assert.equal(parsed.provider, "chatgpt-web");
-  assert.equal(parsed.model, "gpt-5.3-instant");
+  assert.equal(parsed.model, "gpt-5.5");
   const provider = getImageProvider(parsed.provider);
   assert.equal(provider.format, "chatgpt-web");
   assert.equal(provider.authHeader, "cookie");

tests/unit/claude-codex-identity-version-sync.test.ts

@@ -36,12 +36,12 @@ test("Claude CLI version constants are in lockstep across all 4 sources", () =>
   );
 });
 
-test("Claude CLI is pinned to the captured 2.1.195 release", () => {
-  assert.equal(id.CLAUDE_CODE_VERSION, "2.1.195");
+test("Claude CLI is pinned to the captured 2.1.207 release", () => {
+  assert.equal(id.CLAUDE_CODE_VERSION, "2.1.207");
 });
 
-test("Codex client is pinned to the captured 0.144.0 release", () => {
-  assert.equal(codexCfg.getCodexClientVersion(), "0.144.0");
-  assert.equal(codexCfg.getCodexUserAgent(), "codex-cli/0.144.0 (Windows 10.0.26200; x64)");
-  assert.equal(codexCfg.getCodexDefaultHeaders().Version, "0.144.0");
+test("Codex client is pinned to the captured 0.144.1 release", () => {
+  assert.equal(codexCfg.getCodexClientVersion(), "0.144.1");
+  assert.equal(codexCfg.getCodexUserAgent(), "codex-cli/0.144.1 (Windows 10.0.26200; x64)");
+  assert.equal(codexCfg.getCodexDefaultHeaders().Version, "0.144.1");
 });

tests/unit/claude-identity-version-sync.test.ts

@@ -2,7 +2,7 @@ import test from "node:test";
 import assert from "node:assert/strict";
 
 // Claude-Code identity version is hand-bumped in lockstep across several modules
-// (2.1.158 → .187 → .195 …). A silent partial bump makes one surface advertise a stale
+// (2.1.158 → .187 → .195 → .207 …). A silent partial bump makes one surface advertise a stale
 // `claude-cli/<version>` and can break Anthropic identity gating. This guard fails on drift.
 // (quota-share-hardening Phase 2 — gaps v3.8.42.)
 const claudeIdentity = await import("../../open-sse/executors/claudeIdentity.ts");
@@ -13,7 +13,7 @@ const glmProvider = await import("../../open-sse/config/glmProvider.ts");
 
 const CANONICAL = claudeIdentity.CLAUDE_CODE_VERSION;
 
-// "claude-cli/2.1.195 (external, sdk-cli)" → "2.1.195". String ops only — never a RegExp over
+// "claude-cli/2.1.207 (external, sdk-cli)" → "2.1.207". String ops only — never a RegExp over
 // the value, per the project's anti-ReDoS contract.
 function versionFromUserAgent(userAgent: string): string {
   const afterSlash = userAgent.split("claude-cli/")[1] ?? "";

tests/unit/cli-model-config-schema.test.ts

@@ -17,3 +17,20 @@ test("cliModelConfigSchema accepts Codex xhigh reasoning effort", () => {
     assert.equal(result.data.reasoningEffort, "xhigh");
   }
 });
+
+test("cliModelConfigSchema accepts Codex max and ultra reasoning efforts", () => {
+  for (const reasoningEffort of ["max", "ultra"] as const) {
+    const result = cliModelConfigSchema.safeParse({
+      baseUrl: "http://localhost:20128/api/v1",
+      apiKey: "sk_omniroute",
+      model: "gpt-5.6-sol",
+      reasoningEffort,
+      wireApi: "responses",
+    });
+
+    assert.equal(result.success, true, reasoningEffort);
+    if (result.success) {
+      assert.equal(result.data.reasoningEffort, reasoningEffort);
+    }
+  }
+});

tests/unit/cli-tools.test.ts

@@ -86,12 +86,12 @@ test("CLI fingerprint preserves Codex executor User-Agent and maps legacy Copilo
     "codex",
     {
       Authorization: "Bearer token",
-      "User-Agent": "codex-cli/0.144.0 (Windows 10.0.26200; x64)",
+      "User-Agent": "codex-cli/0.144.1 (Windows 10.0.26200; x64)",
     },
     { model: "gpt-5.5", messages: [], stream: true }
   );
 
-  assert.equal(codex.headers["User-Agent"], "codex-cli/0.144.0 (Windows 10.0.26200; x64)");
+  assert.equal(codex.headers["User-Agent"], "codex-cli/0.144.1 (Windows 10.0.26200; x64)");
   assert.deepEqual(Object.keys(JSON.parse(codex.bodyString)), ["model", "stream", "messages"]);
 
   const copilot = applyFingerprint(

tests/unit/client-identity-profiles.test.ts

@@ -17,7 +17,8 @@ const {
   getClientIdentityProfileHeaders,
   isClientIdentityProfileId,
 } = await import("../../src/shared/constants/clientIdentityProfiles.ts");
-const { isForbiddenCustomHeaderName } = await import("../../src/shared/constants/upstreamHeaders.ts");
+const { isForbiddenCustomHeaderName } =
+  await import("../../src/shared/constants/upstreamHeaders.ts");
 const { DefaultExecutor } = await import("../../open-sse/executors/default.ts");
 const core = await import("../../src/lib/db/core.ts");
 
@@ -38,11 +39,11 @@ test("getClientIdentityProfileHeaders: unknown profile id falls back to no heade
 
 test("getClientIdentityProfileHeaders: known CLI profiles expose their preset headers", () => {
   const claudeCli = getClientIdentityProfileHeaders("claude-cli");
-  assert.equal(claudeCli["User-Agent"], "claude-cli/2.1.195 (external, cli)");
+  assert.equal(claudeCli["User-Agent"], "claude-cli/2.1.207 (external, cli)");
   assert.equal(claudeCli["X-App"], "cli");
 
   const codexCli = getClientIdentityProfileHeaders("codex-cli");
-  assert.equal(codexCli["User-Agent"], "codex_cli_rs/0.136.0");
+  assert.equal(codexCli["User-Agent"], "codex_cli_rs/0.144.1");
   assert.equal(codexCli.originator, "codex_cli_rs");
 
   const geminiCli = getClientIdentityProfileHeaders("gemini-cli");
@@ -54,7 +55,7 @@ test("getClientIdentityProfileHeaders: returns a fresh mutable copy (catalog sta
   headers["User-Agent"] = "tampered";
   assert.equal(
     CLIENT_IDENTITY_PROFILES["claude-cli"].headers["User-Agent"],
-    "claude-cli/2.1.195 (external, cli)"
+    "claude-cli/2.1.207 (external, cli)"
   );
 });
 
@@ -79,7 +80,7 @@ test("a selected profile's headers land in providerSpecificData.customHeaders",
     customHeaders: { ...profileHeaders, "X-Operator-Set": "keep-me" },
   };
 
-  assert.equal(providerSpecificData.customHeaders["User-Agent"], "codex_cli_rs/0.136.0");
+  assert.equal(providerSpecificData.customHeaders["User-Agent"], "codex_cli_rs/0.144.1");
   assert.equal(providerSpecificData.customHeaders.originator, "codex_cli_rs");
   assert.equal(providerSpecificData.customHeaders["X-Operator-Set"], "keep-me");
 });
@@ -99,7 +100,7 @@ test("profile headers merged into customHeaders survive applyCustomHeaders sanit
     true
   ) as Record<string, string>;
 
-  assert.equal(headers["User-Agent"], "claude-cli/2.1.195 (external, cli)");
+  assert.equal(headers["User-Agent"], "claude-cli/2.1.207 (external, cli)");
   assert.equal(headers["X-App"], "cli");
   assert.equal(headers["Authorization"], "Bearer test-key");
 });

tests/unit/codex-catalog-revalidation-runtime.test.ts

@@ -0,0 +1,100 @@
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+import test from "node:test";
+
+const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-codex-revalidation-"));
+process.env.DATA_DIR = TEST_DATA_DIR;
+
+const core = await import("../../src/lib/db/core.ts");
+const providersDb = await import("../../src/lib/db/providers.ts");
+const revalidation = await import("../../src/shared/services/codexCatalogRevalidation.ts");
+
+const originalFetch = globalThis.fetch;
+const originalEnv = {
+  OMNIROUTE_PORT: process.env.OMNIROUTE_PORT,
+  PORT: process.env.PORT,
+  DASHBOARD_PORT: process.env.DASHBOARD_PORT,
+  BASE_URL: process.env.BASE_URL,
+  NEXT_PUBLIC_BASE_URL: process.env.NEXT_PUBLIC_BASE_URL,
+  NEXT_PUBLIC_APP_URL: process.env.NEXT_PUBLIC_APP_URL,
+  OMNIROUTE_INTERNAL_SCHEME: process.env.OMNIROUTE_INTERNAL_SCHEME,
+};
+
+async function resetStorage() {
+  globalThis.fetch = originalFetch;
+  core.resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+  fs.mkdirSync(TEST_DATA_DIR, { recursive: true });
+}
+
+test.beforeEach(async () => {
+  await resetStorage();
+  process.env.OMNIROUTE_PORT = "20128";
+  process.env.PORT = "22128";
+  process.env.DASHBOARD_PORT = "22128";
+  process.env.BASE_URL = "https://attacker.example";
+  delete process.env.NEXT_PUBLIC_BASE_URL;
+  delete process.env.NEXT_PUBLIC_APP_URL;
+  delete process.env.OMNIROUTE_INTERNAL_SCHEME;
+});
+
+test.after(() => {
+  globalThis.fetch = originalFetch;
+  core.resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+  for (const [key, value] of Object.entries(originalEnv)) {
+    if (value === undefined) delete process.env[key];
+    else process.env[key] = value;
+  }
+});
+
+test("live Codex revalidation sends its internal header only to the dashboard loopback origin", async () => {
+  const connection = await providersDb.createProviderConnection({
+    provider: "codex",
+    authType: "oauth",
+    name: "Codex Runtime Safety",
+    accessToken: "test-token",
+    isActive: true,
+    providerSpecificData: { workspaceId: "runtime-safety" },
+  });
+  const calls: Array<{ url: string; hasInternalAuth: boolean; redirect?: RequestRedirect }> = [];
+  globalThis.fetch = async (input, init) => {
+    const headers = new Headers(init?.headers);
+    calls.push({
+      url: String(input),
+      hasInternalAuth: headers.has("x-model-sync-internal-auth"),
+      redirect: init?.redirect,
+    });
+    return Response.json({ syncedModels: 1 });
+  };
+
+  const result = await revalidation.liveResyncCodexConnections("http://127.0.0.1:7777");
+
+  assert.deepEqual(result, { attempted: 1, succeeded: 1 });
+  assert.deepEqual(calls, [
+    {
+      url: `http://127.0.0.1:22128/api/providers/${connection.id}/sync-models?quiet=1`,
+      hasInternalAuth: true,
+      redirect: "error",
+    },
+  ]);
+});
+
+test("Codex readiness and live sync resolve the same dashboard loopback port", async () => {
+  const calls: string[] = [];
+  globalThis.fetch = async (input, init) => {
+    calls.push(String(input));
+    assert.equal(init?.redirect, "error");
+    return new Response(null, { status: 404 });
+  };
+
+  await revalidation.waitForLoopbackHttpReady({
+    apiBaseUrl: "http://localhost:7777",
+    maxWaitMs: 100,
+    pollMs: 1,
+  });
+
+  assert.deepEqual(calls, ["http://127.0.0.1:22128/api/providers/__readiness_probe__/models"]);
+});

tests/unit/codex-catalog-revalidation.test.ts

@@ -0,0 +1,257 @@
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+import test from "node:test";
+
+import type { SyncedAvailableModel } from "../../src/lib/db/models.ts";
+import {
+  createCodexCatalogRevalidationCoordinator,
+  executeCodexCatalogRevalidation,
+  resolveBootRevalidationReason,
+  resolveCodexCatalogAppVersion,
+  scrubSyncedModelsWithCodexDenylist,
+} from "../../src/shared/services/codexCatalogRevalidation.ts";
+
+test("Codex revalidation avoids top-level createRequire in packaged Next modules", () => {
+  const source = fs.readFileSync(
+    path.join(process.cwd(), "src/shared/services/codexCatalogRevalidation.ts"),
+    "utf8"
+  );
+  assert.doesNotMatch(source, /^const\s+\w+\s*=\s*createRequire\s*\(/m);
+});
+
+test("scrubSyncedModelsWithCodexDenylist drops the GPT-5.4 family and keeps others", () => {
+  const input = [
+    { id: "gpt-5.6-sol", name: "Sol", source: "imported" },
+    { id: "gpt-5.4", name: "Retired", source: "imported" },
+    { id: "gpt-5.4-mini", name: "Retired Mini", source: "imported" },
+    { id: "future-codex-experimental", name: "Future", source: "imported" },
+  ] satisfies SyncedAvailableModel[];
+  const { kept, removedIds } = scrubSyncedModelsWithCodexDenylist(input);
+
+  assert.deepEqual(
+    kept.map((m) => m.id),
+    ["gpt-5.6-sol", "future-codex-experimental"]
+  );
+  assert.deepEqual(removedIds.sort(), ["gpt-5.4", "gpt-5.4-mini"]);
+});
+
+test("scrubSyncedModelsWithCodexDenylist is a no-op when nothing is denylisted", () => {
+  const input = [
+    { id: "gpt-5.6-sol", name: "Sol", source: "imported" },
+    { id: "gpt-5.5-low", name: "5.5 Low", source: "imported" },
+  ] satisfies SyncedAvailableModel[];
+  const { kept, removedIds } = scrubSyncedModelsWithCodexDenylist(input);
+  assert.equal(removedIds.length, 0);
+  assert.equal(kept.length, 2);
+});
+
+test("resolveCodexCatalogAppVersion uses stable, source-qualified identities", () => {
+  const runtimeRoot = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-version-identity-"));
+  assert.equal(
+    resolveCodexCatalogAppVersion(
+      {
+        OMNIROUTE_BUILD_SHA: "abc123",
+        npm_package_version: "9.9.9",
+      },
+      { runtimeRoot }
+    ),
+    "build:abc123"
+  );
+  assert.equal(
+    resolveCodexCatalogAppVersion(
+      {
+        npm_package_version: "3.8.47",
+      },
+      { runtimeRoot }
+    ),
+    "pkg:3.8.47"
+  );
+
+  try {
+    fs.writeFileSync(path.join(runtimeRoot, "BUILD_SHA"), "sentinel-sha\n");
+    assert.equal(
+      resolveCodexCatalogAppVersion({}, { runtimeRoot, packageVersion: "3.8.47" }),
+      "build:sentinel-sha"
+    );
+    fs.rmSync(path.join(runtimeRoot, "BUILD_SHA"));
+    fs.writeFileSync(path.join(runtimeRoot, "package.json"), '{"version":"9.8.7"}\n');
+    assert.equal(resolveCodexCatalogAppVersion({}, { runtimeRoot }), "pkg:9.8.7");
+    assert.equal(
+      resolveCodexCatalogAppVersion({}, { runtimeRoot, packageVersion: "3.8.47" }),
+      "pkg:3.8.47"
+    );
+    assert.equal(resolveCodexCatalogAppVersion({}, { runtimeRoot, packageVersion: null }), null);
+  } finally {
+    fs.rmSync(runtimeRoot, { recursive: true, force: true });
+  }
+});
+
+test("resolveBootRevalidationReason only fires on first-start or upgrade", () => {
+  assert.equal(resolveBootRevalidationReason(null, "v2"), "first-start");
+  assert.equal(resolveBootRevalidationReason("", "v2"), "first-start");
+  assert.equal(resolveBootRevalidationReason("v1", "v2"), "upgrade");
+  assert.equal(resolveBootRevalidationReason("v2", "v2"), null);
+});
+
+test("executeCodexCatalogRevalidation records completion only after a full live sync", async () => {
+  const events: string[] = [];
+  const failed = await executeCodexCatalogRevalidation({
+    appVersion: "build:audit",
+    scrub: async () => {
+      events.push("scrub");
+    },
+    waitForReady: async () => {
+      events.push("ready");
+    },
+    liveResync: async () => {
+      events.push("sync");
+      return { attempted: 2, succeeded: 1 };
+    },
+    writeMarker: async () => {
+      events.push("marker");
+      return true;
+    },
+    logSuccess: () => {
+      events.push("log");
+    },
+  });
+
+  assert.equal(failed.complete, false);
+  assert.deepEqual(events, ["scrub", "ready", "sync"]);
+
+  events.length = 0;
+  const complete = await executeCodexCatalogRevalidation({
+    appVersion: "build:audit",
+    scrub: async () => {
+      events.push("scrub");
+    },
+    waitForReady: async () => {
+      events.push("ready");
+    },
+    liveResync: async () => {
+      events.push("sync");
+      return { attempted: 2, succeeded: 2 };
+    },
+    writeMarker: async () => {
+      events.push("marker");
+      return true;
+    },
+    logSuccess: () => {
+      events.push("log");
+    },
+  });
+
+  assert.equal(complete.complete, true);
+  assert.deepEqual(events, ["scrub", "ready", "sync", "marker", "log"]);
+});
+
+test("executeCodexCatalogRevalidation leaves an unknown-version run incomplete and unlogged", async () => {
+  const events: string[] = [];
+  const result = await executeCodexCatalogRevalidation({
+    appVersion: null,
+    scrub: async () => undefined,
+    waitForReady: async () => undefined,
+    liveResync: async () => ({ attempted: 0, succeeded: 0 }),
+    writeMarker: async () => {
+      events.push("marker");
+      return true;
+    },
+    logSuccess: () => {
+      events.push("log");
+    },
+  });
+
+  assert.equal(result.complete, false);
+  assert.deepEqual(events, []);
+});
+
+test("executeCodexCatalogRevalidation does not complete after readiness or marker failure", async () => {
+  let liveCalls = 0;
+  let markerCalls = 0;
+  let successLogs = 0;
+  const notReady = await executeCodexCatalogRevalidation({
+    appVersion: "build:audit",
+    scrub: async () => undefined,
+    waitForReady: async () => {
+      throw new Error("not ready");
+    },
+    liveResync: async () => {
+      liveCalls += 1;
+      return { attempted: 1, succeeded: 1 };
+    },
+    writeMarker: async () => {
+      markerCalls += 1;
+      return true;
+    },
+    logSuccess: () => {
+      successLogs += 1;
+    },
+  });
+  assert.equal(notReady.complete, false);
+  assert.equal(liveCalls, 0);
+  assert.equal(markerCalls, 0);
+  assert.equal(successLogs, 0);
+
+  const markerFailed = await executeCodexCatalogRevalidation({
+    appVersion: "build:audit",
+    scrub: async () => undefined,
+    waitForReady: async () => undefined,
+    liveResync: async () => ({ attempted: 1, succeeded: 1 }),
+    writeMarker: async () => false,
+    logSuccess: () => {
+      successLogs += 1;
+    },
+  });
+  assert.equal(markerFailed.complete, false);
+  assert.equal(successLogs, 0);
+});
+
+test("Codex revalidation coordinator coalesces startup and queues one init rerun", async () => {
+  let releaseFirst: (() => void) | undefined;
+  const firstGate = new Promise<void>((resolve) => {
+    releaseFirst = resolve;
+  });
+  const reasons: string[] = [];
+  let active = 0;
+  let maxActive = 0;
+  const request = createCodexCatalogRevalidationCoordinator(async (options) => {
+    reasons.push(options.reason);
+    active += 1;
+    maxActive = Math.max(maxActive, active);
+    if (reasons.length === 1) await firstGate;
+    active -= 1;
+  });
+
+  const startupA = request({ reason: "upgrade" });
+  const startupB = request({ reason: "upgrade" });
+  const initA = request({ reason: "init" });
+  const initB = request({ reason: "init" });
+  releaseFirst?.();
+  await Promise.all([startupA, startupB, initA, initB]);
+
+  assert.deepEqual(reasons, ["upgrade", "init"]);
+  assert.equal(maxActive, 1);
+});
+
+test("Codex revalidation coordinator does not lose init during active-run settlement", async () => {
+  let releaseFirst: (() => void) | undefined;
+  const firstGate = new Promise<void>((resolve) => {
+    releaseFirst = resolve;
+  });
+  const reasons: string[] = [];
+  const request = createCodexCatalogRevalidationCoordinator(async ({ reason }) => {
+    reasons.push(reason);
+    if (reasons.length === 1) await firstGate;
+  });
+
+  const activeRun = request({ reason: "upgrade" });
+  const settlementInit = firstGate.then(() =>
+    Promise.resolve().then(() => request({ reason: "init" }))
+  );
+  releaseFirst?.();
+  await Promise.all([activeRun, settlementInit]);
+
+  assert.deepEqual(reasons, ["upgrade", "init"]);
+});

tests/unit/codex-fast-tier.test.ts

@@ -38,7 +38,11 @@ test("Codex global service mode distinguishes no setting from explicit tiers", (
   );
   assert.deepEqual(
     resolveCodexGlobalFastServiceTier({ codexServiceTier: { enabled: true, tier: "default" } }),
-    { enabled: true, tier: "default", supportedModels: ["gpt-5.5", "gpt-5.4"] }
+    {
+      enabled: true,
+      tier: "default",
+      supportedModels: ["gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna", "gpt-5.5"],
+    }
   );
 });
 

tests/unit/codex-gpt56-catalog.test.ts

@@ -0,0 +1,62 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { getModelsByProviderId } from "../../open-sse/config/providerModels.ts";
+
+test("Codex catalog exposes the GPT-5.6 lineup in configured priority order", () => {
+  const models = getModelsByProviderId("codex");
+  const expectedIds = [
+    "gpt-5.6-sol",
+    "gpt-5.6-sol-ultra",
+    "gpt-5.6-sol-max",
+    "gpt-5.6-sol-xhigh",
+    "gpt-5.6-sol-high",
+    "gpt-5.6-sol-medium",
+    "gpt-5.6-sol-low",
+    "gpt-5.6-terra",
+    "gpt-5.6-terra-ultra",
+    "gpt-5.6-terra-max",
+    "gpt-5.6-terra-xhigh",
+    "gpt-5.6-terra-high",
+    "gpt-5.6-terra-medium",
+    "gpt-5.6-terra-low",
+    "gpt-5.6-luna",
+    "gpt-5.6-luna-max",
+    "gpt-5.6-luna-xhigh",
+    "gpt-5.6-luna-high",
+    "gpt-5.6-luna-medium",
+    "gpt-5.6-luna-low",
+  ];
+
+  assert.deepEqual(
+    models.slice(0, expectedIds.length).map((model) => model.id),
+    expectedIds
+  );
+
+  for (const modelId of expectedIds) {
+    const model = models.find((entry) => entry.id === modelId);
+    assert.ok(model, `codex must expose ${modelId}`);
+    assert.equal(model.contextLength, 500000);
+    assert.equal(model.maxInputTokens, 372000);
+    assert.equal(model.maxOutputTokens, 128000);
+    assert.equal(model.targetFormat, "openai-responses");
+    assert.equal(model.toolCalling, true);
+    assert.equal(model.supportsReasoning, true);
+    assert.equal(model.supportsVision, true);
+    assert.equal(model.supportsXHighEffort, true);
+  }
+
+  assert.equal(
+    models.some((model) => model.id === "gpt-5.6-luna-ultra"),
+    false
+  );
+});
+
+test("Codex catalog no longer exposes GPT-5.4 models", () => {
+  const models = getModelsByProviderId("codex");
+
+  assert.deepEqual(
+    models.filter((model) => model.id.startsWith("gpt-5.4")).map((model) => model.id),
+    []
+  );
+});

tests/unit/codex-models-catalog-refresh.test.ts

@@ -32,8 +32,14 @@ process.env.API_KEY_SECRET = process.env.API_KEY_SECRET || "catalog-test-secret"
 
 const core = await import("../../src/lib/db/core.ts");
 const apiKeysDb = await import("../../src/lib/db/apiKeys.ts");
+const modelsDb = await import("../../src/lib/db/models.ts");
+const providersDb = await import("../../src/lib/db/providers.ts");
 const v1ModelsCatalog = await import("../../src/app/api/v1/models/catalog.ts");
 
+type CatalogResponse = {
+  data?: Array<{ id: string }>;
+};
+
 async function resetStorage() {
   core.resetDbInstance();
   apiKeysDb.resetApiKeyState();
@@ -56,7 +62,8 @@ test("codex client (originator: codex_exec) receives a top-level `models` array
     new Request("http://localhost/v1/models?client_version=0.137.0", {
       headers: {
         originator: "codex_exec",
-        "user-agent": "codex_exec/0.137.0 (Ubuntu 24.4.0; x86_64) vscode/3.7.19 (codex_exec; 0.137.0)",
+        "user-agent":
+          "codex_exec/0.137.0 (Ubuntu 24.4.0; x86_64) vscode/3.7.19 (codex_exec; 0.137.0)",
       },
     })
   );
@@ -103,3 +110,48 @@ test("non-codex OpenAI client keeps the unchanged {object,data} shape (no `model
     "non-codex clients must NOT receive a `models` key (response stays byte-identical)"
   );
 });
+
+test("v1 models catalog exposes remote-only Codex IDs from the discovery cache", async () => {
+  const connection = await providersDb.createProviderConnection({
+    provider: "codex",
+    authType: "oauth",
+    name: "codex-curated-catalog",
+    accessToken: "codex-access-token",
+    isActive: true,
+    testStatus: "active",
+  });
+
+  await modelsDb.replaceSyncedAvailableModelsForConnection("codex", connection.id, [
+    {
+      id: "codex-auto-review",
+      name: "Codex Auto Review Remote",
+      source: "imported",
+      supportedEndpoints: ["responses"],
+    },
+    {
+      id: "future-codex-model",
+      name: "Future Codex Model",
+      source: "imported",
+      supportedEndpoints: ["responses"],
+    },
+    {
+      id: "gpt-5.4-mini",
+      name: "Retired GPT-5.4 Mini",
+      source: "imported",
+      supportedEndpoints: ["responses"],
+    },
+  ]);
+
+  const response = await v1ModelsCatalog.getUnifiedModelsResponse(
+    new Request("http://localhost/api/v1/models")
+  );
+  const body = (await response.json()) as CatalogResponse;
+  const ids = new Set((body.data || []).map((item) => item.id));
+
+  assert.equal(response.status, 200);
+  assert.equal(ids.has("cx/codex-auto-review"), true);
+  assert.equal(ids.has("cx/future-codex-model"), true);
+  assert.equal(ids.has("codex/future-codex-model"), true);
+  assert.equal(ids.has("cx/gpt-5.4-mini"), false);
+  assert.equal(ids.has("codex/gpt-5.4-mini"), false);
+});

tests/unit/codex-stream-false.test.ts

@@ -213,14 +213,14 @@ test.after(async () => {
 test("CodexExecutor.transformRequest clones the request body before forcing stream=true", () => {
   const executor = new CodexExecutor();
   const body = {
-    model: "gpt-5.4",
+    model: "gpt-5.6-sol",
     input: [{ role: "user", content: [{ type: "input_text", text: "Oi" }] }],
     stream: false,
     reasoning: { effort: "low" },
   };
   const original = structuredClone(body);
 
-  const transformed = executor.transformRequest("gpt-5.4", body, false, {
+  const transformed = executor.transformRequest("gpt-5.6-sol", body, false, {
     requestEndpointPath: "/responses",
   });
 
@@ -314,7 +314,7 @@ test("chatCore converts Responses-style NDJSON fallback into JSON when stream=fa
 test("handleComboChat validates non-stream quality using the original client stream intent", async () => {
   const combo = {
     name: "codex-stream-false-quality",
-    models: ["codex/gpt-5.4", "openai/gpt-4o-mini"],
+    models: ["codex/gpt-5.6-sol", "openai/gpt-4o-mini"],
   };
   const log = createComboLog();
   const seenModels = [];
@@ -327,7 +327,7 @@ test("handleComboChat validates non-stream quality using the original client str
     combo,
     handleSingleModel: async (requestBody, modelStr) => {
       seenModels.push(modelStr);
-      if (modelStr === "codex/gpt-5.4") {
+      if (modelStr === "codex/gpt-5.6-sol") {
         requestBody.stream = true;
         return jsonResponse({
           choices: [
@@ -355,7 +355,7 @@ test("handleComboChat validates non-stream quality using the original client str
   const payload = (await result.json()) as any;
 
   assert.equal(result.ok, true);
-  assert.deepEqual(seenModels, ["codex/gpt-5.4", "openai/gpt-4o-mini"]);
+  assert.deepEqual(seenModels, ["codex/gpt-5.6-sol", "openai/gpt-4o-mini"]);
   assert.equal(payload.choices[0].message.content, "Brasilia");
   assert.ok(
     log.entries.some(

tests/unit/combo-body-specific-400-stop-4279.test.ts

@@ -68,7 +68,7 @@ test("#4279 combo stops at the first body-specific 400 instead of trying every t
 
   const result = await handleComboChat({
     body: { model: "test", messages: [{ role: "user", content: "hi" }] },
-    combo: makeCombo(["codex/gpt-5.2", "codex/gpt-5.3-codex", "codex/gpt-5.4"]),
+    combo: makeCombo(["codex/gpt-5.2", "codex/gpt-5.3-codex", "codex/gpt-5.6-sol"]),
     handleSingleModel,
     log,
     settings: {},

tests/unit/combo-config.test.ts

@@ -369,7 +369,7 @@ test("resolveComboConfig tolerates invalid or missing inputs and falls back to d
 test("createComboSchema accepts context-relay strategy with handoff config", () => {
   const parsed = createComboSchema.parse({
     name: "codex-relay",
-    models: ["codex/gpt-5.4"],
+    models: ["codex/gpt-5.6-sol"],
     strategy: "context-relay",
     config: {
       handoffThreshold: 0.85,
@@ -443,7 +443,7 @@ test("createComboSchema accepts structured combo steps with pinned connection an
         kind: "model",
         id: "step-codex-a",
         providerId: "codex",
-        model: "gpt-5.4",
+        model: "gpt-5.6-sol",
         connectionId: "conn-codex-a",
         weight: 10,
       },
@@ -472,14 +472,14 @@ test("createComboSchema accepts composite tiers that reference normalized combo
         kind: "model",
         id: "step-primary",
         providerId: "codex",
-        model: "gpt-5.4",
+        model: "gpt-5.6-sol",
         connectionId: "conn-codex-a",
       },
       {
         kind: "model",
         id: "step-backup",
         providerId: "codex",
-        model: "gpt-5.4",
+        model: "gpt-5.6-sol",
         connectionId: "conn-codex-b",
       },
     ],

tests/unit/combo-context-relay.test.ts

@@ -156,14 +156,14 @@ test("handleComboChat context-relay skips unavailable models and falls through t
     combo: {
       name: "relay-skip-unavailable",
       strategy: "context-relay",
-      models: ["codex/gpt-5.4", "openai/gpt-4o-mini"],
+      models: ["codex/gpt-5.6-sol", "openai/gpt-4o-mini"],
       config: { maxRetries: 0 },
     },
     handleSingleModel: async (_body, modelStr) => {
       calls.push(modelStr);
       return okResponse();
     },
-    isModelAvailable: async (modelStr) => modelStr !== "codex/gpt-5.4",
+    isModelAvailable: async (modelStr) => modelStr !== "codex/gpt-5.6-sol",
     log: createLog(),
     settings: null,
     allCombos: null,
@@ -181,7 +181,7 @@ test("handleComboChat context-relay treats provider circuit breaker responses as
   const combo = {
     name: "relay-breaker",
     strategy: "context-relay",
-    models: ["codex/gpt-5.4", "openai/gpt-4o-mini"],
+    models: ["codex/gpt-5.6-sol", "openai/gpt-4o-mini"],
     config: { maxRetries: 0 },
   };
   const calls = [];
@@ -193,7 +193,7 @@ test("handleComboChat context-relay treats provider circuit breaker responses as
     combo,
     handleSingleModel: async (_body, modelStr) => {
       calls.push(modelStr);
-      if (modelStr === "codex/gpt-5.4") {
+      if (modelStr === "codex/gpt-5.6-sol") {
         return providerBreakerOpenResponse();
       }
       return okResponse();
@@ -205,7 +205,7 @@ test("handleComboChat context-relay treats provider circuit breaker responses as
   });
 
   assert.equal(result.ok, true);
-  assert.deepEqual(calls, ["codex/gpt-5.4", "openai/gpt-4o-mini"]);
+  assert.deepEqual(calls, ["codex/gpt-5.6-sol", "openai/gpt-4o-mini"]);
 });
 
 test("handleComboChat context-relay persists a handoff when codex quota reaches the warning threshold", async () => {
@@ -235,7 +235,7 @@ test("handleComboChat context-relay persists a handoff when codex quota reaches
     combo: {
       name: "relay-generate",
       strategy: "context-relay",
-      models: ["codex/gpt-5.4"],
+      models: ["codex/gpt-5.6-sol"],
       config: { maxRetries: 0, handoffThreshold: 0.85, handoffProviders: ["codex"] },
     },
     handleSingleModel: async (body) => {
@@ -309,7 +309,7 @@ test("handleComboChat context-relay respects handoffProviders and skips generati
     combo: {
       name: "relay-disabled-provider",
       strategy: "context-relay",
-      models: ["codex/gpt-5.4"],
+      models: ["codex/gpt-5.6-sol"],
       config: { maxRetries: 0, handoffProviders: ["openai"] },
     },
     handleSingleModel: async (body) => {
@@ -363,7 +363,7 @@ test("handleComboChat context-relay treats explicit empty handoffProviders as di
     combo: {
       name: "relay-empty-providers",
       strategy: "context-relay",
-      models: ["codex/gpt-5.4"],
+      models: ["codex/gpt-5.6-sol"],
       config: { maxRetries: 0, handoffProviders: [] },
     },
     handleSingleModel: async () => okResponse(),

tests/unit/context-handoff.test.ts

@@ -46,7 +46,7 @@ test("buildHandoffSystemMessage and injectHandoffIntoBody preserve existing hist
     taskProgress: "Need to finish tests",
     activeEntities: ["combo.ts", "chat.ts"],
     messageCount: 42,
-    model: "codex/gpt-5.4",
+    model: "codex/gpt-5.6-sol",
     warningThresholdPct: 0.85,
     generatedAt: "2099-04-08T12:00:00.000Z",
     expiresAt: "2099-04-08T17:00:00.000Z",
@@ -79,7 +79,7 @@ test("injectHandoffIntoBody preserves Responses API shape for native Codex reque
     taskProgress: "Need to carry state across account switches",
     activeEntities: ["chat.ts", "contextHandoff.ts"],
     messageCount: 8,
-    model: "codex/gpt-5.4",
+    model: "codex/gpt-5.6-sol",
     warningThresholdPct: 0.85,
     generatedAt: "2099-04-08T12:00:00.000Z",
     expiresAt: "2099-04-08T17:00:00.000Z",
@@ -139,7 +139,7 @@ test("maybeGenerateHandoff skips below the warning threshold", async () => {
     connectionId: "conn-low",
     percentUsed: 0.7,
     messages: [{ role: "user", content: "hello" }],
-    model: "codex/gpt-5.4",
+    model: "codex/gpt-5.6-sol",
     expiresAt: null,
     handleSingleModel: async () => {
       called = true;
@@ -164,7 +164,7 @@ test("maybeGenerateHandoff persists a structured handoff once the threshold is r
       { role: "user", content: "Please continue wiring the combo" },
       { role: "assistant", content: "Working on it" },
     ],
-    model: "codex/gpt-5.4",
+    model: "codex/gpt-5.6-sol",
     expiresAt: "2099-04-08T17:00:00.000Z",
     handleSingleModel: async (body, modelStr) => {
       calls.push({ body, modelStr });
@@ -197,7 +197,7 @@ test("maybeGenerateHandoff persists a structured handoff once the threshold is r
   assert.equal(saved.summary, "Relay summary generated");
   assert.deepEqual(saved.keyDecisions, ["Use context-relay"]);
   assert.equal(calls.length, 1);
-  assert.equal(calls[0].modelStr, "codex/gpt-5.4");
+  assert.equal(calls[0].modelStr, "codex/gpt-5.6-sol");
   assert.equal(calls[0].body._omnirouteSkipContextRelay, true);
   assert.equal(calls[0].body._omnirouteInternalRequest, "context-handoff");
 });
@@ -215,7 +215,7 @@ test("maybeGenerateHandoff deduplicates concurrent in-flight generations for the
     connectionId: "conn-dedupe",
     percentUsed: 0.89,
     messages: [{ role: "user", content: "Generate once" }],
-    model: "codex/gpt-5.4",
+    model: "codex/gpt-5.6-sol",
     expiresAt: "2099-01-01T00:00:00.000Z",
     handleSingleModel: async () => {
       calls.push("summary");
@@ -265,7 +265,7 @@ test("maybeGenerateHandoff allows a new attempt after a failed in-flight generat
     connectionId: "conn-retry",
     percentUsed: 0.9,
     messages: [{ role: "user", content: "Retry after failure" }],
-    model: "codex/gpt-5.4",
+    model: "codex/gpt-5.6-sol",
     expiresAt: "2099-01-01T00:00:00.000Z",
     handleSingleModel: async () => {
       calls += 1;
@@ -316,7 +316,7 @@ test("maybeGenerateHandoff respects explicit empty handoffProviders and skips ge
     connectionId: "conn-disabled",
     percentUsed: 0.92,
     messages: [{ role: "user", content: "Do not generate" }],
-    model: "codex/gpt-5.4",
+    model: "codex/gpt-5.6-sol",
     expiresAt: null,
     config: { handoffProviders: [] },
     handleSingleModel: async () => {
@@ -340,7 +340,7 @@ test("context handoff DB module upserts and deletes active handoffs", () => {
     taskProgress: "step one",
     activeEntities: ["a.ts"],
     messageCount: 3,
-    model: "codex/gpt-5.4",
+    model: "codex/gpt-5.6-sol",
     warningThresholdPct: 0.85,
     generatedAt: "2099-04-08T10:00:00.000Z",
     expiresAt: "2099-01-01T00:00:00.000Z",
@@ -354,7 +354,7 @@ test("context handoff DB module upserts and deletes active handoffs", () => {
     taskProgress: "step two",
     activeEntities: ["b.ts"],
     messageCount: 4,
-    model: "codex/gpt-5.4",
+    model: "codex/gpt-5.6-sol",
     warningThresholdPct: 0.86,
     generatedAt: "2099-04-08T11:00:00.000Z",
     expiresAt: "2099-01-01T00:00:00.000Z",
@@ -385,7 +385,7 @@ test("selectMessagesForSummary filters falsy values and preserves system/develop
     messages as contextHandoff.MessageLike[],
     2
   );
-  
+
   assert.equal(selected.length, 4);
   assert.equal(selected[0].role, "system");
   assert.equal(selected[1].role, "developer");

tests/unit/executor-codex-gpt56.test.ts

@@ -0,0 +1,52 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { CodexExecutor } from "../../open-sse/executors/codex.ts";
+
+test("CodexExecutor.transformRequest preserves max effort for GPT-5.6", () => {
+  const executor = new CodexExecutor();
+  const result = executor.transformRequest(
+    "gpt-5.6-sol",
+    {
+      model: "gpt-5.6-sol",
+      input: [],
+      reasoning_effort: "max",
+    },
+    false,
+    { requestEndpointPath: "/responses" }
+  );
+
+  assert.equal(result.model, "gpt-5.6-sol");
+  assert.equal(result.reasoning.effort, "max");
+  assert.equal(result.reasoning_effort, undefined);
+});
+
+test("CodexExecutor.transformRequest maps GPT-5.6 ultra aliases to max wire effort", () => {
+  const executor = new CodexExecutor();
+
+  for (const model of ["gpt-5.6-sol-ultra", "gpt-5.6-terra-ultra"]) {
+    const result = executor.transformRequest(model, { model, input: [] }, false, {
+      requestEndpointPath: "/responses",
+    });
+
+    assert.equal(result.model, model.replace(/-ultra$/, ""));
+    assert.equal(result.reasoning.effort, "max");
+  }
+});
+
+test("CodexExecutor.transformRequest clamps Luna ultra requests to its max effort", () => {
+  const executor = new CodexExecutor();
+  const result = executor.transformRequest(
+    "gpt-5.6-luna",
+    {
+      model: "gpt-5.6-luna",
+      input: [],
+      reasoning_effort: "ultra",
+    },
+    false,
+    { requestEndpointPath: "/responses" }
+  );
+
+  assert.equal(result.model, "gpt-5.6-luna");
+  assert.equal(result.reasoning.effort, "max");
+});

tests/unit/executor-codex.test.ts

@@ -86,16 +86,15 @@ test("Codex helper functions isolate rate-limit scopes and parse quota headers",
   assert.equal(getCodexModelScope("gpt-5.5-xhigh"), "codex");
   assert.equal(getCodexUpstreamModel("gpt-5.5-xhigh"), "gpt-5.5");
   assert.equal(getCodexUpstreamModel("gpt-5.5-medium"), "gpt-5.5");
+  assert.equal(getCodexUpstreamModel("gpt-5.1-codex-max"), "gpt-5.1-codex-max");
   // With mock WS transport + codexTransport=websocket, gpt-5.5 models require WS
-  __setCodexWebSocketTransportForTesting(
-    async (): Promise<MockCodexWebSocket> => ({
-      send() {},
-      close() {},
-      onmessage: null,
-      onerror: null,
-      onclose: null,
-    })
-  );
+  __setCodexWebSocketTransportForTesting(async (): Promise<MockCodexWebSocket> => ({
+    send() {},
+    close() {},
+    onmessage: null,
+    onerror: null,
+    onclose: null,
+  }));
   assert.equal(
     isCodexResponsesWebSocketRequired("gpt-5.5-xhigh", {
       providerSpecificData: { codexTransport: "websocket" },
@@ -185,10 +184,10 @@ test("CodexExecutor.buildHeaders binds workspace ids and disables SSE accept for
   assert.equal(standardHeaders.Authorization, "Bearer codex-token");
   assert.equal(standardHeaders.Accept, "text/event-stream");
   assert.equal(standardHeaders["chatgpt-account-id"], "workspace-1");
-  assert.equal(standardHeaders.Version, "0.144.0");
+  assert.equal(standardHeaders.Version, "0.144.1");
   assert.equal(standardHeaders["Openai-Beta"], "responses=experimental");
   assert.equal(standardHeaders["X-Codex-Beta-Features"], "responses_websockets");
-  assert.equal(standardHeaders["User-Agent"], "codex-cli/0.144.0 (Windows 10.0.26200; x64)");
+  assert.equal(standardHeaders["User-Agent"], "codex-cli/0.144.1 (Windows 10.0.26200; x64)");
   assert.equal(compactHeaders.Accept, "application/json");
 });
 
@@ -214,7 +213,7 @@ test("CodexExecutor.buildHeaders honors safe env overrides for Version and User-
     },
     () => {
       const headers = executor.buildHeaders({ accessToken: "codex-token" }, true);
-      assert.equal(headers.Version, "0.144.0");
+      assert.equal(headers.Version, "0.144.1");
       assert.equal(headers["User-Agent"], "custom-codex/9.9.9");
     }
   );
@@ -799,12 +798,12 @@ test("CodexExecutor.transformRequest keeps GPT 5.3 Codex reasoning in Responses
   assert.equal(sanitized.reasoning_effort, undefined);
 });
 
-test("CodexExecutor.transformRequest passes GPT 5.4 Mini xhigh reasoning through unchanged in Responses shape (#3756)", () => {
+test("CodexExecutor.transformRequest passes GPT 5.6 Luna xhigh reasoning through unchanged", () => {
   const executor = new CodexExecutor();
   const transformed = executor.transformRequest(
-    "gpt-5.4-mini",
+    "gpt-5.6-luna",
     {
-      model: "gpt-5.4-mini",
+      model: "gpt-5.6-luna",
       input: [],
       reasoning: { effort: "xhigh", summary: "detailed" },
       include: ["code_interpreter_call.outputs"],
@@ -817,12 +816,12 @@ test("CodexExecutor.transformRequest passes GPT 5.4 Mini xhigh reasoning through
   const sanitized = sanitizeReasoningEffortForProvider(
     transformed,
     "codex",
-    "gpt-5.4-mini",
+    "gpt-5.6-luna",
     null
   ) as Record<string, unknown>;
   const reasoning = getRecord(sanitized.reasoning);
 
-  assert.equal(sanitized.model, "gpt-5.4-mini");
+  assert.equal(sanitized.model, "gpt-5.6-luna");
   assert.deepEqual(reasoning, { effort: "xhigh", summary: "detailed" });
   assert.deepEqual(sanitized.include, [
     "code_interpreter_call.outputs",
@@ -1058,9 +1057,9 @@ test("CodexExecutor.execute skips identity headers for unsafe session ids", asyn
 test("CodexExecutor.transformRequest preserves namespace MCP tools and hosted tool types", () => {
   const executor = new CodexExecutor();
   const result = executor.transformRequest(
-    "gpt-5.4",
+    "gpt-5.6-sol",
     {
-      model: "gpt-5.4",
+      model: "gpt-5.6-sol",
       input: [],
       tools: [
         { type: "function", name: "exec_command", parameters: { type: "object" } },

tests/unit/glm-executor.test.ts

@@ -181,7 +181,7 @@ test("GlmExecutor separates OpenAI-compatible coding headers from Anthropic head
   assert.equal(anthropicHeaders["anthropic-version"], "2023-06-01");
   assert.match(anthropicHeaders["anthropic-beta"], /claude-code-20250219/);
   assert.equal(anthropicHeaders["anthropic-dangerous-direct-browser-access"], "true");
-  assert.match(anthropicHeaders["User-Agent"], /^claude-cli\/2\.1\.195 \(external, sdk-cli\)$/);
+  assert.match(anthropicHeaders["User-Agent"], /^claude-cli\/2\.1\.207 \(external, sdk-cli\)$/);
   assert.equal(anthropicHeaders["X-Stainless-Lang"], "js");
   assert.equal(anthropicHeaders["X-Stainless-Runtime"], "node");
 });

tests/unit/gpt-max-input-tokens-6191.test.ts

@@ -55,10 +55,10 @@ test("all codex gpt-5.5 effort variants carry the distinct input cap (#6191)", (
 });
 
 test("regression: a model without maxInputTokens still falls back to its context window", () => {
-  // codex gpt-5.4 declares no maxInputTokens, so max_input_tokens must equal
-  // the context window (the historical fallback) — no under-reporting.
-  const caps = modelCapabilities.getResolvedModelCapabilities("codex/gpt-5.4");
-  assert.ok((caps.contextWindow ?? 0) > 0, "gpt-5.4 should have a context window");
+  // OpenAI GPT-5.4 declares a context window without maxInputTokens, so the
+  // historical fallback must still avoid under-reporting.
+  const caps = modelCapabilities.getResolvedModelCapabilities("openai/gpt-5.4");
+  assert.ok((caps.contextWindow ?? 0) > 0, "OpenAI GPT-5.4 should have a context window");
   assert.equal(
     caps.maxInputTokens,
     caps.contextWindow,

tests/unit/gpt5-sampling-guard.test.ts

@@ -61,8 +61,8 @@ test("model suffix -high triggers strip; -none keeps sampling", () => {
 });
 
 test("non-openai provider is untouched (codex is guarded by the executor allowlist)", () => {
-  const body = { model: "gpt-5.4", temperature: 0.7, reasoning_effort: "high" };
-  const result = stripGpt5SamplingWhenReasoning(body, "codex", "gpt-5.4");
+  const body = { model: "gpt-5.6-sol", temperature: 0.7, reasoning_effort: "high" };
+  const result = stripGpt5SamplingWhenReasoning(body, "codex", "gpt-5.6-sol");
   assert.equal(result.temperature, 0.7);
 });
 

tests/unit/guide-settings-route.test.ts

@@ -224,9 +224,9 @@ test("guide-settings POST preserves existing OpenCode config fields while only u
     body: JSON.stringify({
       baseUrl: "http://my-omni/v1",
       apiKey: "sk-123",
-      models: ["cx/gpt-5.4", "opencode-go/kimi-k2.6"],
+      models: ["cx/gpt-5.6-sol", "opencode-go/kimi-k2.6"],
       modelLabels: {
-        "cx/gpt-5.4": "GPT-5.4",
+        "cx/gpt-5.6-sol": "GPT-5.6 Sol",
         "opencode-go/kimi-k2.6": "Kimi K2.6",
       },
     }),
@@ -252,7 +252,7 @@ test("guide-settings POST preserves existing OpenCode config fields while only u
   assert.equal(content.provider.omniroute.options.baseURL, "http://my-omni/v1");
   assert.ok(content.provider.omniroute.options.apiKey.startsWith("sk-"));
   assert.deepEqual(content.provider.omniroute.models, {
-    "cx/gpt-5.4": { name: "GPT-5.4" },
+    "cx/gpt-5.6-sol": { name: "GPT-5.6 Sol" },
     "opencode-go/kimi-k2.6": { name: "Kimi K2.6" },
   });
 });

tests/unit/image-generation-handler.test.ts

@@ -1814,7 +1814,7 @@ test("handleImageGeneration routes codex image requests through /responses with
   try {
     const result = await handleImageGeneration({
       body: {
-        model: "codex/gpt-5.4",
+        model: "codex/gpt-5.6-sol",
         prompt: "Draw a happy red kitten",
         response_format: "b64_json",
       },
@@ -1829,7 +1829,7 @@ test("handleImageGeneration routes codex image requests through /responses with
     assert.equal(captured.url, "https://chatgpt.com/backend-api/codex/responses");
     assert.equal(captured.headers.Authorization, "Bearer codex-token");
     assert.equal(captured.headers["chatgpt-account-id"], "acct-123");
-    assert.equal(captured.body.model, "gpt-5.4");
+    assert.equal(captured.body.model, "gpt-5.6-sol");
     assert.equal(captured.body.stream, true);
     assert.equal(captured.body.store, false);
     assert.deepEqual(captured.body.tools, [{ type: "image_generation", output_format: "png" }]);
@@ -1853,7 +1853,7 @@ test("handleImageGeneration (codex) returns a data URL when response_format is n
 
   try {
     const result = await handleImageGeneration({
-      body: { model: "cx/gpt-5.4", prompt: "kitten" },
+      body: { model: "cx/gpt-5.6-sol", prompt: "kitten" },
       credentials: { accessToken: "codex-token" },
       log: null,
     });
@@ -1895,7 +1895,7 @@ test("handleImageGeneration (codex) fans out n>1 requests in parallel", async ()
   try {
     pending = handleImageGeneration({
       body: {
-        model: "codex/gpt-5.4",
+        model: "codex/gpt-5.6-sol",
         prompt: "kitten",
         n: 2,
         response_format: "b64_json",
@@ -1937,7 +1937,7 @@ test("handleImageGeneration (codex) surfaces an error when no image_generation_c
 
   try {
     const result = await handleImageGeneration({
-      body: { model: "codex/gpt-5.4", prompt: "kitten" },
+      body: { model: "codex/gpt-5.6-sol", prompt: "kitten" },
       credentials: { accessToken: "codex-token" },
       log: null,
     });
@@ -1956,7 +1956,7 @@ test("handleImageGeneration (codex) propagates upstream HTTP errors", async () =
 
   try {
     const result = await handleImageGeneration({
-      body: { model: "codex/gpt-5.4", prompt: "kitten" },
+      body: { model: "codex/gpt-5.6-sol", prompt: "kitten" },
       credentials: { accessToken: "codex-token" },
       log: null,
     });
@@ -1982,7 +1982,7 @@ test("handleImageGeneration (codex) forwards size and maps GPT-Image quality to
   try {
     await handleImageGeneration({
       body: {
-        model: "codex/gpt-5.4",
+        model: "codex/gpt-5.6-sol",
         prompt: "kitten",
         size: "1024x1792",
         quality: "hd",
@@ -2000,14 +2000,14 @@ test("handleImageGeneration (codex) forwards size and maps GPT-Image quality to
     ]);
 
     await handleImageGeneration({
-      body: { model: "codex/gpt-5.4", prompt: "kitten", quality: "standard" },
+      body: { model: "codex/gpt-5.6-sol", prompt: "kitten", quality: "standard" },
       credentials: { accessToken: "codex-token" },
       log: null,
     });
     assert.equal(captured.tools[0].quality, "medium");
 
     await handleImageGeneration({
-      body: { model: "codex/gpt-5.4", prompt: "kitten" },
+      body: { model: "codex/gpt-5.6-sol", prompt: "kitten" },
       credentials: { accessToken: "codex-token" },
       log: null,
     });

tests/unit/image-generation-route.test.ts

@@ -69,7 +69,7 @@ test("v1 image models GET exposes image-only modalities for credential-backed im
   assert.deepEqual((byId.get("stability-ai/fast") as any).input_modalities, ["image"]);
 });
 
-test("v1 image models GET hides providers without active credentials", async () => {
+test("v1 image models GET exposes current Codex image models and hides inactive providers", async () => {
   await seedConnection("codex", { apiKey: "codex-key" });
 
   const response = await imageRoute.GET();
@@ -77,7 +77,11 @@ test("v1 image models GET hides providers without active credentials", async ()
   const ids = body.data.map((item) => item.id);
 
   assert.equal(response.status, 200);
-  assert.ok(ids.includes("codex/gpt-5.5"));
+  assert.deepEqual(
+    ids.filter((id) => id.startsWith("codex/")),
+    ["codex/gpt-5.6-sol", "codex/gpt-5.6-terra", "codex/gpt-5.6-luna"]
+  );
+  assert.ok(!ids.includes("codex/gpt-5.5"));
   assert.ok(!ids.includes("openai/gpt-image-2"));
   assert.ok(!ids.some((id: string) => id.startsWith("xai/")));
 });
@@ -147,7 +151,7 @@ test("v1 image edit POST enforces disabled API key policy", async () => {
 
   const formData = new FormData();
   formData.set("prompt", "make the background lighter");
-  formData.set("model", "cgpt-web/gpt-5.3-instant");
+  formData.set("model", "cgpt-web/gpt-5.5");
   formData.set("image", new File([new Uint8Array([1, 2, 3])], "source.png", { type: "image/png" }));
 
   const response = await imageEditRoute.POST(

tests/unit/image-registry-gpt56.test.ts

@@ -0,0 +1,26 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { IMAGE_PROVIDERS, parseImageModel } from "../../open-sse/config/imageRegistry.ts";
+
+test("ChatGPT Web image catalog exposes GPT-5.5 Instant instead of GPT-5.3 Instant", () => {
+  assert.deepEqual(IMAGE_PROVIDERS["chatgpt-web"].models, [
+    { id: "gpt-5.5", name: "GPT-5.5 Instant (ChatGPT Web Image)" },
+  ]);
+  assert.deepEqual(parseImageModel("cgpt-web/gpt-5.5"), {
+    provider: "chatgpt-web",
+    model: "gpt-5.5",
+  });
+});
+
+test("Codex image catalog exposes only the GPT-5.6 Sol, Terra, and Luna models", () => {
+  assert.deepEqual(IMAGE_PROVIDERS.codex.models, [
+    { id: "gpt-5.6-sol", name: "GPT 5.6 Sol (Codex Image)" },
+    { id: "gpt-5.6-terra", name: "GPT 5.6 Terra (Codex Image)" },
+    { id: "gpt-5.6-luna", name: "GPT 5.6 Luna (Codex Image)" },
+  ]);
+
+  for (const model of ["gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna"]) {
+    assert.deepEqual(parseImageModel(`cx/${model}`), { provider: "codex", model });
+  }
+});

tests/unit/image-routes-combo-edits-3214-3215.test.ts

@@ -124,16 +124,17 @@ test("resolveImageRouteModel lets bare combos shadow built-in image aliases", as
 });
 
 test("resolveImageRouteModel keeps codex bare aliases over same-name combos", async () => {
-  await createCombo({ name: "gpt-5.5", models: ["myimg/gpt-5.5"], strategy: "priority" });
+  await createCombo({
+    name: "gpt-5.6-sol",
+    models: ["myimg/gpt-5.6-sol"],
+    strategy: "priority",
+  });
 
-  assert.equal(await resolveSingleImageComboTarget("gpt-5.5"), "myimg/gpt-5.5");
-  assert.equal(await resolveImageRouteModel("gpt-5.5"), "gpt-5.5");
+  assert.equal(await resolveSingleImageComboTarget("gpt-5.6-sol"), "myimg/gpt-5.6-sol");
+  assert.equal(await resolveImageRouteModel("gpt-5.6-sol"), "gpt-5.6-sol");
 });
 
 test("resolveImageRouteModel leaves built-in / already-resolved ids untouched", async () => {
-  assert.equal(
-    await resolveImageRouteModel("cgpt-web/gpt-5.3-instant"),
-    "cgpt-web/gpt-5.3-instant"
-  );
+  assert.equal(await resolveImageRouteModel("cgpt-web/gpt-5.5"), "cgpt-web/gpt-5.5");
   assert.equal(await resolveSingleImageComboTarget("definitely-not-a-combo-3215"), null);
 });

tests/unit/model-alias-provider-resolution.test.ts

@@ -19,82 +19,82 @@ import assert from "node:assert/strict";
 
 import { getModelInfoCore } from "../../open-sse/services/model.ts";
 
-// ── Scenario A: direct exact alias "gpt-5.4" → "cx/gpt-5.4" ─────────────────
-// When aliases contain { "gpt-5.4": "cx/gpt-5.4" }, the request for "gpt-5.4"
+// ── Scenario A: direct exact alias "gpt-5.6-sol" → "cx/gpt-5.6-sol" ─────────
+// When aliases contain { "gpt-5.6-sol": "cx/gpt-5.6-sol" }, the request for "gpt-5.6-sol"
 // must resolve to provider "codex" (cx = codex alias), NOT "openai".
 
-test("A: custom alias gpt-5.4 → cx/gpt-5.4 resolves to codex, not openai", async () => {
-  const aliases = { "gpt-5.4": "cx/gpt-5.4" };
-  const result = await getModelInfoCore("gpt-5.4", aliases);
+test("A: custom alias gpt-5.6-sol → cx/gpt-5.6-sol resolves to codex, not openai", async () => {
+  const aliases = { "gpt-5.6-sol": "cx/gpt-5.6-sol" };
+  const result = await getModelInfoCore("gpt-5.6-sol", aliases);
   assert.strictEqual(
     result.provider,
     "codex",
     `Expected provider "codex" but got "${result.provider}" — alias was overridden by provider inference`
   );
-  assert.strictEqual(result.model, "gpt-5.4");
+  assert.strictEqual(result.model, "gpt-5.6-sol");
 });
 
-test("A: custom alias gpt-5.4 → cx/gpt-5.4 with async getter also resolves to codex", async () => {
-  const aliases = { "gpt-5.4": "cx/gpt-5.4" };
-  const result = await getModelInfoCore("gpt-5.4", async () => aliases);
+test("A: custom alias gpt-5.6-sol → cx/gpt-5.6-sol with async getter resolves to codex", async () => {
+  const aliases = { "gpt-5.6-sol": "cx/gpt-5.6-sol" };
+  const result = await getModelInfoCore("gpt-5.6-sol", async () => aliases);
   assert.strictEqual(result.provider, "codex");
-  assert.strictEqual(result.model, "gpt-5.4");
+  assert.strictEqual(result.model, "gpt-5.6-sol");
 });
 
-test("A: without alias, gpt-5.4 still resolves to openai via inference (baseline)", async () => {
+test("A: without alias, gpt-5.6-sol resolves to openai via inference (baseline)", async () => {
   // Ensure that inference still works when no alias is configured.
-  const result = await getModelInfoCore("gpt-5.4", {});
+  const result = await getModelInfoCore("gpt-5.6-sol", {});
   assert.strictEqual(result.provider, "openai");
-  assert.strictEqual(result.model, "gpt-5.4");
+  assert.strictEqual(result.model, "gpt-5.6-sol");
 });
 
-// ── Scenario B: wildcard alias "*gpt-5.4*" → "cx/gpt-5.4" ───────────────────
+// ── Scenario B: wildcard alias "*gpt-5.6*" → "cx/gpt-5.6-sol" ───────────────
 // Glob patterns are also supported via resolveWildcardAlias.
 
-test("B: wildcard alias *gpt-5.4* → cx/gpt-5.4 resolves to codex", async () => {
-  const aliases = { "*gpt-5.4*": "cx/gpt-5.4" };
-  const result = await getModelInfoCore("gpt-5.4", aliases);
+test("B: wildcard alias *gpt-5.6* → cx/gpt-5.6-sol resolves to codex", async () => {
+  const aliases = { "*gpt-5.6*": "cx/gpt-5.6-sol" };
+  const result = await getModelInfoCore("gpt-5.6-sol", aliases);
   assert.strictEqual(result.provider, "codex");
-  assert.strictEqual(result.model, "gpt-5.4");
+  assert.strictEqual(result.model, "gpt-5.6-sol");
 });
 
-test("B: wildcard alias *gpt-5* → cx/gpt-5.4 resolves to codex for gpt-5.4", async () => {
-  const aliases = { "*gpt-5*": "cx/gpt-5.4" };
-  const result = await getModelInfoCore("gpt-5.4", aliases);
+test("B: wildcard alias *gpt-5* → cx/gpt-5.6-sol resolves to codex", async () => {
+  const aliases = { "*gpt-5*": "cx/gpt-5.6-sol" };
+  const result = await getModelInfoCore("gpt-5.6-sol", aliases);
   assert.strictEqual(result.provider, "codex");
 });
 
-// ── Scenario C: explicit provider prefix "openai/gpt-5.4" ────────────────────
-// When the client sends "openai/gpt-5.4" explicitly, parseModel returns
+// ── Scenario C: explicit provider prefix "openai/gpt-5.6-sol" ────────────────
+// When the client sends "openai/gpt-5.6-sol" explicitly, parseModel returns
 // isAlias=false. This bypasses the alias lookup in getModelInfoCore (by design —
-// explicit provider prefixes override aliases). The alias "gpt-5.4 → cx/gpt-5.4"
+// explicit provider prefixes override aliases). The alias "gpt-5.6-sol → cx/gpt-5.6-sol"
 // should NOT apply here; this is intentional behavior.
-// NOTE: Aliases for "openai/gpt-5.4" → "cx/gpt-5.4" are a distinct key and
+// NOTE: Aliases for "openai/gpt-5.6-sol" → "cx/gpt-5.6-sol" are a distinct key and
 // must be configured explicitly if desired.
 
-test("C: explicit openai/gpt-5.4 resolves to openai regardless of bare alias", async () => {
-  // The alias is for bare "gpt-5.4", NOT for "openai/gpt-5.4".
+test("C: explicit openai/gpt-5.6-sol resolves to openai regardless of bare alias", async () => {
+  // The alias is for bare "gpt-5.6-sol", NOT for "openai/gpt-5.6-sol".
   // An explicit provider prefix takes precedence over aliases on the bare name.
-  const aliases = { "gpt-5.4": "cx/gpt-5.4" };
-  const result = await getModelInfoCore("openai/gpt-5.4", aliases);
+  const aliases = { "gpt-5.6-sol": "cx/gpt-5.6-sol" };
+  const result = await getModelInfoCore("openai/gpt-5.6-sol", aliases);
   assert.strictEqual(result.provider, "openai");
-  assert.strictEqual(result.model, "gpt-5.4");
+  assert.strictEqual(result.model, "gpt-5.6-sol");
 });
 
-test("C: explicit provider prefix alias openai/gpt-5.4 → cx/gpt-5.4 is NOT applied (isAlias=false path)", async () => {
+test("C: explicit openai/gpt-5.6-sol alias is not applied on the isAlias=false path", async () => {
   // A slashful input is NOT treated as an alias key by getModelInfoCore.
   // The alias lookup only runs when isAlias=true (bare model name).
-  // Users wanting to override openai/gpt-5.4 must configure cx/gpt-5.4 explicitly.
-  const aliases = { "openai/gpt-5.4": "cx/gpt-5.4" };
-  const result = await getModelInfoCore("openai/gpt-5.4", aliases);
-  // The slashful input is parsed as provider=openai, model=gpt-5.4 — alias not consulted
+  // Users wanting Codex must send cx/gpt-5.6-sol explicitly.
+  const aliases = { "openai/gpt-5.6-sol": "cx/gpt-5.6-sol" };
+  const result = await getModelInfoCore("openai/gpt-5.6-sol", aliases);
+  // The slashful input is parsed as provider=openai, model=gpt-5.6-sol — alias not consulted.
   assert.strictEqual(result.provider, "openai");
 });
 
 // ── Scenario D: non-gpt aliases are unaffected ───────────────────────────────
 
 test("D: unrelated models with no alias still route by inference", async () => {
-  const aliases = { "gpt-5.4": "cx/gpt-5.4" };
+  const aliases = { "gpt-5.6-sol": "cx/gpt-5.6-sol" };
   const result = await getModelInfoCore("gpt-4o", aliases);
   assert.strictEqual(result.provider, "openai");
   assert.strictEqual(result.model, "gpt-4o");
@@ -102,7 +102,10 @@ test("D: unrelated models with no alias still route by inference", async () => {
 
 test("D: claude model alias routes to requested provider", async () => {
   // Explicit alias for a multi-provider model wins over inference
-  const aliases = { "gpt-5.4": "cx/gpt-5.4", "my-claude": "anthropic/claude-opus-4-7" };
+  const aliases = {
+    "gpt-5.6-sol": "cx/gpt-5.6-sol",
+    "my-claude": "anthropic/claude-opus-4-7",
+  };
   const result = await getModelInfoCore("my-claude", aliases);
   assert.strictEqual(result.provider, "anthropic");
   assert.strictEqual(result.model, "claude-opus-4-7");

tests/unit/model-parse.test.ts

@@ -22,10 +22,10 @@ test("[1m] suffix: works with provider prefix", () => {
 });
 
 test("parseModel trims provider prefix and model id", () => {
-  const result = parseModel("  cx / gpt-5.4  ");
+  const result = parseModel("  cx / gpt-5.6-sol  ");
   assert.strictEqual(result.providerAlias, "cx");
   assert.strictEqual(result.provider, "codex");
-  assert.strictEqual(result.model, "gpt-5.4");
+  assert.strictEqual(result.model, "gpt-5.6-sol");
 });
 
 test("parseModel treats exact slashful model ids as models, not provider prefixes", () => {

tests/unit/model-sync-custom-preservation.test.ts

@@ -0,0 +1,78 @@
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+import test from "node:test";
+
+const TEST_DATA_DIR = fs.mkdtempSync(
+  path.join(os.tmpdir(), "omniroute-model-sync-custom-preservation-")
+);
+process.env.DATA_DIR = TEST_DATA_DIR;
+process.env.API_KEY_SECRET ||= `test-model-sync-custom-${Date.now()}`;
+
+const core = await import("../../src/lib/db/core.ts");
+const providersDb = await import("../../src/lib/db/providers.ts");
+const modelsDb = await import("../../src/lib/db/models.ts");
+const modelSyncRoute = await import("../../src/app/api/providers/[id]/sync-models/route.ts");
+const scheduler = await import("../../src/shared/services/modelSyncScheduler.ts");
+
+const originalFetch = globalThis.fetch;
+
+test.after(() => {
+  globalThis.fetch = originalFetch;
+  core.resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+});
+
+test("model sync preserves response-only custom models during discovery", async () => {
+  const connection = await providersDb.createProviderConnection({
+    provider: "codex",
+    authType: "oauth",
+    name: "Codex Custom Preservation",
+    accessToken: "test-codex-token",
+    providerSpecificData: { workspaceId: "workspace-custom-preservation" },
+  });
+  await modelsDb.addCustomModel(
+    "codex",
+    "operator-private-codex",
+    "Operator Private Codex",
+    "manual",
+    "responses",
+    ["responses"],
+    "openai-responses",
+    { inputTokenLimit: 123456, outputTokenLimit: 6543 }
+  );
+  const before = await modelsDb.getCustomModels("codex");
+
+  globalThis.fetch = async (input) => {
+    const url = new URL(String(input));
+    if (url.pathname.includes("__readiness_probe__")) {
+      return new Response(null, { status: 404 });
+    }
+    if (url.pathname === `/api/providers/${connection.id}/models`) {
+      const models = [{ id: "future-codex-experimental", name: "Future Codex" }];
+      if (url.searchParams.get("excludeCustom") !== "true") {
+        models.push({ id: "operator-private-codex", name: "Operator Private Codex" });
+      }
+      return Response.json({ models, source: "api" });
+    }
+    throw new Error(`Unexpected fetch in custom preservation test: ${url.href}`);
+  };
+
+  const response = await modelSyncRoute.POST(
+    new Request(`http://localhost/api/providers/${connection.id}/sync-models?quiet=1`, {
+      method: "POST",
+      headers: scheduler.buildModelSyncInternalHeaders(),
+    }),
+    { params: { id: connection.id } }
+  );
+
+  assert.equal(response.status, 200);
+  assert.deepEqual(await modelsDb.getCustomModels("codex"), before);
+  assert.deepEqual(
+    (await modelsDb.getSyncedAvailableModelsForConnection("codex", connection.id)).map(
+      (model) => model.id
+    ),
+    ["future-codex-experimental"]
+  );
+});

tests/unit/model-sync-route.test.ts

@@ -74,7 +74,7 @@ test("model sync route skips success log when fetched models do not change store
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return Response.json({
       models: [{ id: "custom-model-1", name: "Custom Model 1" }],
@@ -119,7 +119,7 @@ test("model sync route stores the real provider while keeping the account label"
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return Response.json({
       models: [{ id: "custom-model-2", name: "Custom Model 2" }],
@@ -204,7 +204,7 @@ test("model sync route propagates upstream failures and records an error log ent
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return Response.json({ error: "Provider upstream unavailable" }, { status: 502 });
   };
@@ -241,7 +241,7 @@ test("model sync route falls back to the upstream HTTP status when the models pa
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return Response.json({}, { status: 429 });
   };
@@ -277,7 +277,7 @@ test("model sync route reports invalid JSON /models responses without losing ups
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return new Response("<html>bad gateway</html>", {
       status: 200,
@@ -325,7 +325,7 @@ test("model sync route preserves previously synced models when the upstream omit
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return Response.json({});
   };
@@ -369,7 +369,7 @@ test("model sync route writes synced available models for Gemini connections", a
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return Response.json({
       models: [
@@ -433,7 +433,7 @@ test("model sync route writes synced available models for non-Gemini providers t
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return Response.json({
       models: [
@@ -489,7 +489,7 @@ test("model sync route import mode merges discovered models without deleting man
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return Response.json({
       models: [{ id: "router-v4", name: "Router V4" }],
@@ -555,7 +555,7 @@ test("model sync route import mode ignores supported endpoint ordering changes",
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return Response.json({
       models: [
@@ -619,7 +619,7 @@ test("model sync route import mode reports updates without counting them as new
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return Response.json({
       models: [
@@ -693,7 +693,7 @@ test("model sync route records added, removed, and updated model diffs with fall
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return Response.json({
       models: [
@@ -772,7 +772,7 @@ test("model sync route forwards cookies, filters built-ins, and syncs aliases fo
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     assert.equal(init.headers.cookie, "session=test-cookie");
     assert.equal(
@@ -837,7 +837,7 @@ test("model sync route reports synced managed models separately from preserved m
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return Response.json({
       models: [{ id: "router-v4", name: "Router V4" }],
@@ -902,7 +902,7 @@ test("model sync route uses provider-node prefixes when syncing compatible-provi
     if (String(url).includes("__readiness_probe__")) return new Response(null, { status: 404 });
     assert.equal(
       String(url),
-      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true`
+      `http://127.0.0.1:20128/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`
     );
     return Response.json({
       models: [{ id: "sonnet-4-6", name: "Sonnet 4.6" }],
@@ -957,7 +957,7 @@ test("model sync route falls back to in-process discovery when internal self-fet
 
     fetchCalls.push(urlString);
 
-    if (urlString === `http://localhost/api/providers/${connection.id}/models?refresh=true`) {
+    if (urlString.includes("/models?refresh=true&excludeCustom=true")) {
       throw new Error("fetch failed");
     }
 
@@ -1004,7 +1004,7 @@ test("model sync route falls back to in-process discovery when internal self-fet
   // Route forces IPv4 origin (http://127.0.0.1:PORT) — never "localhost" — to avoid
   // ::1 (IPv6) resolution issues in containers. PORT defaults to 20128 when env unset.
   const expectedPort = process.env.OMNIROUTE_PORT || process.env.PORT || "20128";
-  const selfFetchUrl = `http://127.0.0.1:${expectedPort}/api/providers/${connection.id}/models?refresh=true`;
+  const selfFetchUrl = `http://127.0.0.1:${expectedPort}/api/providers/${connection.id}/models?refresh=true&excludeCustom=true`;
   assert.equal(
     fetchCalls.slice(0, 3).every((u) => u === selfFetchUrl),
     true,

tests/unit/model-sync-scheduler.test.ts

@@ -130,6 +130,157 @@ test("modelSyncScheduler: internal auth headers validate only for scheduler requ
   assert.equal(isModelSyncInternalRequest(externalRequest), false);
 });
 
+test("modelSyncScheduler resolves only loopback origins and uses the dashboard port", async () => {
+  const previous = {
+    OMNIROUTE_PORT: process.env.OMNIROUTE_PORT,
+    PORT: process.env.PORT,
+    DASHBOARD_PORT: process.env.DASHBOARD_PORT,
+    BASE_URL: process.env.BASE_URL,
+    NEXT_PUBLIC_BASE_URL: process.env.NEXT_PUBLIC_BASE_URL,
+    NEXT_PUBLIC_APP_URL: process.env.NEXT_PUBLIC_APP_URL,
+    OMNIROUTE_BASE_PATH: process.env.OMNIROUTE_BASE_PATH,
+    OMNIROUTE_INTERNAL_SCHEME: process.env.OMNIROUTE_INTERNAL_SCHEME,
+    OMNIROUTE_TLS_CERT: process.env.OMNIROUTE_TLS_CERT,
+    OMNIROUTE_TLS_KEY: process.env.OMNIROUTE_TLS_KEY,
+  };
+  process.env.OMNIROUTE_PORT = "20128";
+  process.env.PORT = "22128";
+  process.env.DASHBOARD_PORT = "22128";
+  process.env.BASE_URL = "https://attacker.example";
+  delete process.env.NEXT_PUBLIC_BASE_URL;
+  delete process.env.NEXT_PUBLIC_APP_URL;
+  process.env.OMNIROUTE_BASE_PATH = "/omniroute/";
+  delete process.env.OMNIROUTE_INTERNAL_SCHEME;
+  delete process.env.OMNIROUTE_TLS_CERT;
+  delete process.env.OMNIROUTE_TLS_KEY;
+
+  try {
+    const scheduler = await loadScheduler("trusted-loopback-origin");
+    assert.equal(scheduler.getModelSyncInternalBaseUrl(), "http://127.0.0.1:22128/omniroute");
+    assert.equal(
+      scheduler.resolveModelSyncInternalBaseUrl("https://attacker.example/steal"),
+      "http://127.0.0.1:22128/omniroute"
+    );
+    assert.equal(
+      scheduler.resolveModelSyncInternalBaseUrl("http://127.0.0.1:7777/nested/path"),
+      "http://127.0.0.1:22128/omniroute"
+    );
+    assert.equal(
+      scheduler.resolveModelSyncInternalBaseUrl("http://0.0.0.0:7777/nested/path"),
+      "http://127.0.0.1:22128/omniroute"
+    );
+    assert.equal(
+      scheduler.resolveModelSyncInternalBaseUrl("http://user:pass@localhost:7777"),
+      "http://127.0.0.1:22128/omniroute"
+    );
+  } finally {
+    for (const [key, value] of Object.entries(previous)) {
+      if (value === undefined) delete process.env[key];
+      else process.env[key] = value;
+    }
+  }
+});
+
+test("modelSyncScheduler does not infer the internal listener scheme from public URLs", async () => {
+  const previous = {
+    DASHBOARD_PORT: process.env.DASHBOARD_PORT,
+    BASE_URL: process.env.BASE_URL,
+    NEXT_PUBLIC_BASE_URL: process.env.NEXT_PUBLIC_BASE_URL,
+    NEXT_PUBLIC_APP_URL: process.env.NEXT_PUBLIC_APP_URL,
+    OMNIROUTE_INTERNAL_SCHEME: process.env.OMNIROUTE_INTERNAL_SCHEME,
+    OMNIROUTE_TLS_CERT: process.env.OMNIROUTE_TLS_CERT,
+    OMNIROUTE_TLS_KEY: process.env.OMNIROUTE_TLS_KEY,
+  };
+  process.env.DASHBOARD_PORT = "22128";
+  process.env.BASE_URL = "https://attacker.example";
+  process.env.NEXT_PUBLIC_BASE_URL = "file:///tmp/not-http";
+  process.env.NEXT_PUBLIC_APP_URL = "https://localhost:7777/ignored";
+  delete process.env.OMNIROUTE_INTERNAL_SCHEME;
+  delete process.env.OMNIROUTE_TLS_CERT;
+  delete process.env.OMNIROUTE_TLS_KEY;
+
+  try {
+    const scheduler = await loadScheduler("safe-loopback-fallback");
+    assert.equal(scheduler.getModelSyncInternalBaseUrl(), "http://127.0.0.1:22128");
+  } finally {
+    for (const [key, value] of Object.entries(previous)) {
+      if (value === undefined) delete process.env[key];
+      else process.env[key] = value;
+    }
+  }
+});
+
+test("modelSyncScheduler uses the listener-declared TLS scheme without trusting candidates", async () => {
+  const previous = {
+    DASHBOARD_PORT: process.env.DASHBOARD_PORT,
+    BASE_URL: process.env.BASE_URL,
+    NEXT_PUBLIC_BASE_URL: process.env.NEXT_PUBLIC_BASE_URL,
+    NEXT_PUBLIC_APP_URL: process.env.NEXT_PUBLIC_APP_URL,
+    OMNIROUTE_BASE_PATH: process.env.OMNIROUTE_BASE_PATH,
+    OMNIROUTE_INTERNAL_SCHEME: process.env.OMNIROUTE_INTERNAL_SCHEME,
+    OMNIROUTE_TLS_CERT: process.env.OMNIROUTE_TLS_CERT,
+    OMNIROUTE_TLS_KEY: process.env.OMNIROUTE_TLS_KEY,
+  };
+  process.env.DASHBOARD_PORT = "22128";
+  process.env.BASE_URL = "https://attacker.example";
+  delete process.env.NEXT_PUBLIC_BASE_URL;
+  delete process.env.NEXT_PUBLIC_APP_URL;
+  process.env.OMNIROUTE_BASE_PATH = "/omniroute";
+  process.env.OMNIROUTE_INTERNAL_SCHEME = "https";
+  delete process.env.OMNIROUTE_TLS_CERT;
+  delete process.env.OMNIROUTE_TLS_KEY;
+
+  try {
+    const scheduler = await loadScheduler("trusted-native-tls");
+    assert.equal(
+      scheduler.resolveModelSyncInternalBaseUrl("https://attacker.example:7777/steal"),
+      "https://localhost:22128/omniroute"
+    );
+    assert.equal(
+      scheduler.resolveModelSyncInternalBaseUrl("https://127.0.0.1:7777/nested/path"),
+      "https://localhost:22128/omniroute"
+    );
+  } finally {
+    for (const [key, value] of Object.entries(previous)) {
+      if (value === undefined) delete process.env[key];
+      else process.env[key] = value;
+    }
+  }
+});
+
+test("modelSyncScheduler pins HTTPS transport to IPv4 while retaining localhost SNI", async () => {
+  const scheduler = await loadScheduler("pinned-loopback-tls-connector");
+  let forwardedOptions: Record<string, unknown> | undefined;
+  const connector = scheduler.createPinnedModelSyncTlsConnector((options, callback) => {
+    forwardedOptions = options;
+    callback(new Error("stop before socket creation"), null);
+  });
+
+  connector(
+    {
+      hostname: "localhost",
+      host: "localhost:22128",
+      protocol: "https:",
+      port: "22128",
+    },
+    () => undefined
+  );
+
+  assert.equal(forwardedOptions?.hostname, "127.0.0.1");
+  assert.equal(forwardedOptions?.servername, "localhost");
+});
+
+test("runtime launchers publish the actual internal listener scheme", () => {
+  const runNext = fs.readFileSync(path.join(process.cwd(), "scripts/dev/run-next.mjs"), "utf8");
+  const standalone = fs.readFileSync(
+    path.join(process.cwd(), "scripts/dev/standalone-server-ws.mjs"),
+    "utf8"
+  );
+
+  assert.match(runNext, /OMNIROUTE_INTERNAL_SCHEME\s*=\s*["']http["']/);
+  assert.match(standalone, /OMNIROUTE_INTERNAL_SCHEME\s*=\s*tlsOptions\s*\?\s*["']https["']/);
+});
+
 test("initCloudSync: startup initialization also starts model sync scheduler", () => {
   const filePath = path.join(process.cwd(), "src/lib/initCloudSync.ts");
   const source = fs.readFileSync(filePath, "utf8");
@@ -224,8 +375,10 @@ test("modelSyncScheduler starts once, honors env interval and syncs only active
     await timers.timeouts[0].fn();
 
     assert.equal(fetchCalls.length, 1);
+    assert.match(fetchCalls[0].url, /^http:\/\/127\.0\.0\.1:20128\//);
     assert.match(fetchCalls[0].url, /\/api\/providers\/.*\/sync-models$/);
     assert.equal(fetchCalls[0].options.method, "POST");
+    assert.equal(fetchCalls[0].options.redirect, "error");
     assert.equal(fetchCalls[0].options.headers["Content-Type"], "application/json");
     assert.equal(
       fetchCalls[0].options.headers[scheduler.getModelSyncInternalAuthHeaderName()],