feat(sse): add GLM-5.3 models and effort tiers (#10896)

Merged — clean extraction from #10358's genuinely new content (see PR body for the rationale: an unrelated .planning/codebase/ scaffolding dump was dropped). typecheck/file-size/changelog/provider-consistency gates clean, 18/18 tests passing.

src/shared/constants/modelSpecs.ts

@@ -66,7 +66,14 @@ const BEDROCK_CLAUDE_ALIASES = (...modelIds: string[]) => [
 // Provider discovery/sync sources can under-report GLM-5.2 IDs as 128K.
 // Keep native/bare Z.AI GLM-5.2 context authoritative, but do not blindly apply
 // it to every provider-wrapped alias: hosted providers can and do cap lower.
-const AUTHORITATIVE_CONTEXT_WINDOW_MODEL_IDS = new Set(["glm-5.2", "glm-5.2-high", "glm-5.2-max"]);
+const AUTHORITATIVE_CONTEXT_WINDOW_MODEL_IDS = new Set([
+  "glm-5.3",
+  "glm-5.3-high",
+  "glm-5.3-low",
+  "glm-5.2",
+  "glm-5.2-high",
+  "glm-5.2-max",
+]);
 const AUTHORITATIVE_PROVIDER_CONTEXT_WINDOWS = new Map<string, number>([
   ["cloudflare-ai/@cf/zai-org/glm-5.2", 262144],
   // Hugging Face Router has 1M-capable backends, but bare routing can select
@@ -567,6 +574,30 @@ export const MODEL_SPECS: Record<string, ModelSpec> = {
     supportsTools: true,
   },
 
+  // ── Z.AI GLM-5.3 (1M context mirrored from 5.2 — same base model; 128K max
+  // output; effort via reasoning_effort param, tiers are OmniRoute aliases) ──
+  "glm-5.3": {
+    maxOutputTokens: 131072,
+    contextWindow: 1000000,
+    thinkingBudgetCap: 38912,
+    supportsThinking: true,
+    supportsTools: true,
+  },
+  "glm-5.3-high": {
+    maxOutputTokens: 131072,
+    contextWindow: 1000000,
+    thinkingBudgetCap: 38912,
+    supportsThinking: true,
+    supportsTools: true,
+  },
+  "glm-5.3-low": {
+    maxOutputTokens: 131072,
+    contextWindow: 1000000,
+    thinkingBudgetCap: 38912,
+    supportsThinking: true,
+    supportsTools: true,
+  },
+
   // ── Z.AI GLM-5.2 (1M context, 128K max output, effort tiers) ────
   "glm-5.2": {
     maxOutputTokens: 131072,

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

@@ -111,6 +111,30 @@ export const CLAUDE_SONNET_5_PRICING = {
 };
 
 export const GLM_PRICING = {
+  // GLM-5.3 (2026-08-14): Z.ai hasn't published 5.3 rates yet — mirrored from
+  // GLM-5.2 (same base model; 5.1 and 5.2 also share identical rates).
+  // Correct when https://docs.z.ai/guides/overview/pricing lists glm-5.3.
+  "glm-5.3": {
+    input: 1.2,
+    output: 5,
+    cached: 0.3,
+    reasoning: 5,
+    cache_creation: 1.2,
+  },
+  "glm-5.3-high": {
+    input: 1.2,
+    output: 5,
+    cached: 0.3,
+    reasoning: 5,
+    cache_creation: 1.2,
+  },
+  "glm-5.3-low": {
+    input: 1.2,
+    output: 5,
+    cached: 0.3,
+    reasoning: 5,
+    cache_creation: 1.2,
+  },
   "glm-5.2": {
     input: 1.2,
     output: 5,