diff --git a/CHANGELOG.md b/CHANGELOG.md
index 30123fc466..5c4803349a 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -24,6 +24,7 @@ _In development — bullets added per PR; finalized at release._
 - **fix(providers): the "Test Connection One-by-One" OAuth probe can no longer hang the queue forever** — the OAuth connection-test path called bare `fetch(url, { method, headers })` with no `AbortController`/signal/timeout, so when a provider's probe endpoint accepted the socket but never responded, the awaited fetch never settled and the one-by-one test queue stalled indefinitely (the API-key path was already bounded via `validateProviderApiKey`'s `timeoutMs`). Both the initial probe and the post-refresh retry are now bounded with `AbortSignal.timeout(30s)` — matching the API-key path's 30s budget — and a timed-out probe resolves as a failure with a clear `Test timed out after 30s` message in the same shape as every other test error. (thanks @ntdung6868)
 - **fix(providers): a deactivated account is labeled distinctly from a revoked token** — a Codex connection whose OAuth refresh is fully healthy but whose ChatGPT account has been deactivated by the provider gets a `401` from the upstream API. The connection test labeled that the same as a bad credential (`Token invalid or revoked` → `upstream_auth_error`), so the operator couldn't tell a deactivated account from a revoked token. The test now reads the `401`/`403` body and, when it indicates account deactivation, classifies it as `account_deactivated` — which the dashboard already renders as "Account Deactivated". A plain auth `401` is unchanged. (thanks @ntdung6868)
 - **fix(db): cascade-delete orphaned model aliases when a provider is removed** — deleting a custom provider removed its connections and node but left behind the imported model-alias rows (stored as `key=<alias>`, `value="<providerId>/<model>"`). Those stale aliases then blocked re-importing the same provider — the import dedup treated them as "already exists", so no new models appeared. A new `deleteModelAliasesForProvider(providerId)` DB helper drops every alias whose stored value begins with `<providerId>/` (leaving other providers and user-defined settings aliases untouched), and the provider-node DELETE handler now calls it after removing the connections and node, so a fresh import is unblocked. (thanks @nguyenvanhuy0612)
+- **fix(api): persist `max_input_tokens` / `max_output_tokens` when adding a custom model** — `POST /api/provider-models` silently dropped the per-model token limits set in the "add custom model" form: the handler destructured the rest of the body but never read `max_input_tokens` / `max_output_tokens`, and `addCustomModel()` had no parameter for them, so the values were thrown away on write. The DB layer (`inputTokenLimit` / `outputTokenLimit`) and the `/v1/models` catalog already round-trip these fields — only the write path was missing. The validation schema now accepts the two optional limits, the handler forwards them, and `addCustomModel()` persists them so a custom model's context/output window survives into the catalog. (thanks @codename-zen)
 
 ---
 
diff --git a/src/app/api/provider-models/route.ts b/src/app/api/provider-models/route.ts
index 7856a38b0a..4f982d5912 100644
--- a/src/app/api/provider-models/route.ts
+++ b/src/app/api/provider-models/route.ts
@@ -97,8 +97,18 @@ export async function POST(request) {
     if (isValidationFailure(validation)) {
       return Response.json({ error: validation.error }, { status: 400 });
     }
-    const { provider, modelId, modelName, source, apiFormat, supportedEndpoints, targetFormat } =
-      validation.data;
+    const {
+      provider,
+      modelId,
+      modelName,
+      source,
+      apiFormat,
+      supportedEndpoints,
+      targetFormat,
+      // #1294: persist the per-model token limits set in the add-model form.
+      max_input_tokens: maxInputTokens,
+      max_output_tokens: maxOutputTokens,
+    } = validation.data;
 
     const model = await addCustomModel(
       provider,
@@ -107,7 +117,11 @@ export async function POST(request) {
       source || "manual",
       apiFormat,
       supportedEndpoints,
-      targetFormat
+      targetFormat,
+      {
+        ...(maxInputTokens != null ? { inputTokenLimit: maxInputTokens } : {}),
+        ...(maxOutputTokens != null ? { outputTokenLimit: maxOutputTokens } : {}),
+      }
     );
     return Response.json({ model });
   } catch (error) {
diff --git a/src/lib/db/models.ts b/src/lib/db/models.ts
index a9a7ddd603..35afc39a85 100644
--- a/src/lib/db/models.ts
+++ b/src/lib/db/models.ts
@@ -408,7 +408,10 @@ export async function addCustomModel(
   // #2905: optional per-model wire format override (e.g. "claude" for an
   // opencode-go custom model). When unset, routing falls back to the provider
   // default format.
-  targetFormat?: string
+  targetFormat?: string,
+  // #1294: optional per-model token limits supplied from the "add custom model"
+  // form. Persisted under the same keys the /v1/models catalog reads back.
+  tokenLimits: { inputTokenLimit?: number; outputTokenLimit?: number } = {}
 ) {
   const db = getDbInstance();
   const row = db
@@ -427,6 +430,12 @@ export async function addCustomModel(
     apiFormat,
     supportedEndpoints,
     ...(targetFormat ? { targetFormat } : {}),
+    ...(tokenLimits.inputTokenLimit != null
+      ? { inputTokenLimit: tokenLimits.inputTokenLimit }
+      : {}),
+    ...(tokenLimits.outputTokenLimit != null
+      ? { outputTokenLimit: tokenLimits.outputTokenLimit }
+      : {}),
   };
   models.push(model);
   db.prepare(
diff --git a/src/shared/validation/schemas/provider.ts b/src/shared/validation/schemas/provider.ts
index fd478a0d07..3c89152d82 100644
--- a/src/shared/validation/schemas/provider.ts
+++ b/src/shared/validation/schemas/provider.ts
@@ -405,6 +405,11 @@ export const providerModelMutationSchema = z.object({
   targetFormat: z
     .enum(["openai", "openai-responses", "claude", "gemini", "gemini-cli", "antigravity"])
     .optional(),
+  // #1294: optional token limits set in the "add custom model" form. The wire
+  // shape uses max_input_tokens / max_output_tokens (mirrors the /v1/models
+  // catalog); they persist as inputTokenLimit / outputTokenLimit.
+  max_input_tokens: z.number().int().positive().optional(),
+  max_output_tokens: z.number().int().positive().optional(),
   normalizeToolCallId: z.boolean().optional(),
   preserveOpenAIDeveloperRole: z.boolean().nullable().optional(),
   upstreamHeaders: upstreamHeadersRecordSchema.nullable().optional(),
diff --git a/tests/unit/provider-models-token-limits.test.ts b/tests/unit/provider-models-token-limits.test.ts
new file mode 100644
index 0000000000..aada0b2f87
--- /dev/null
+++ b/tests/unit/provider-models-token-limits.test.ts
@@ -0,0 +1,80 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+const TEST_DATA_DIR = fs.mkdtempSync(
+  path.join(os.tmpdir(), "omniroute-provider-model-token-limits-")
+);
+process.env.DATA_DIR = TEST_DATA_DIR;
+
+const core = await import("../../src/lib/db/core.ts");
+const modelsDb = await import("../../src/lib/db/models.ts");
+const providerModelsRoute = await import("../../src/app/api/provider-models/route.ts");
+
+async function resetStorage() {
+  core.resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+  fs.mkdirSync(TEST_DATA_DIR, { recursive: true });
+}
+
+function buildPostRequest(body) {
+  return new Request("http://localhost/api/provider-models", {
+    method: "POST",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify(body),
+  });
+}
+
+test.beforeEach(async () => {
+  await resetStorage();
+});
+
+test.after(async () => {
+  core.resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+});
+
+// #1294: POST /api/provider-models must persist max_input_tokens / max_output_tokens
+// (stored as inputTokenLimit / outputTokenLimit) so the token limits set in the
+// "add custom model" form survive into the catalog round-trip.
+test("POST persists max_input_tokens / max_output_tokens as inputTokenLimit / outputTokenLimit", async () => {
+  const response = await providerModelsRoute.POST(
+    buildPostRequest({
+      provider: "openai-compatible-demo",
+      modelId: "custom-long-context",
+      modelName: "Custom Long Context",
+      max_input_tokens: 200000,
+      max_output_tokens: 16384,
+    })
+  );
+  const body = (await response.json()) as any;
+
+  assert.equal(response.status, 200);
+  assert.equal(body.model?.id, "custom-long-context");
+  assert.equal(body.model?.inputTokenLimit, 200000);
+  assert.equal(body.model?.outputTokenLimit, 16384);
+
+  const stored = await modelsDb.getCustomModels("openai-compatible-demo");
+  const persisted = stored.find((model) => model.id === "custom-long-context");
+  assert.ok(persisted, "custom model should be persisted");
+  assert.equal(persisted.inputTokenLimit, 200000);
+  assert.equal(persisted.outputTokenLimit, 16384);
+});
+
+test("POST omits token limits when they are not provided", async () => {
+  const response = await providerModelsRoute.POST(
+    buildPostRequest({
+      provider: "openai-compatible-demo",
+      modelId: "custom-no-limits",
+      modelName: "Custom No Limits",
+    })
+  );
+  const body = (await response.json()) as any;
+
+  assert.equal(response.status, 200);
+  assert.equal(body.model?.id, "custom-no-limits");
+  assert.equal("inputTokenLimit" in body.model, false);
+  assert.equal("outputTokenLimit" in body.model, false);
+});