diff --git a/packages/core/src/sync/providers/openrouter.ts b/packages/core/src/sync/providers/openrouter.ts
index 2aeb1e01be..d35cc3f978 100644
--- a/packages/core/src/sync/providers/openrouter.ts
+++ b/packages/core/src/sync/providers/openrouter.ts
@@ -178,7 +178,7 @@ export function buildOpenRouterModel(
   const reasoning_options = existing?.reasoning_options?.length
     ? existing.reasoning_options
     : openRouterReasoningOptions(model.reasoning) ?? existing?.reasoning_options;
-  const context = model.top_provider.context_length ?? model.context_length;
+  const context = model.context_length;
   const family = inferFamily(model, name);
   const releaseDate = dateFromTimestamp(model.created);
   const familyValue = existing?.family === "o" && family !== "o"
diff --git a/packages/core/test/sync.test.ts b/packages/core/test/sync.test.ts
index 8bb40288e0..61bb21792c 100644
--- a/packages/core/test/sync.test.ts
+++ b/packages/core/test/sync.test.ts
@@ -738,6 +738,23 @@ test("syncs OpenRouter reasoning efforts from model metadata", () => {
   });
 });
 
+test("uses OpenRouter model context when top provider reports a shorter context", () => {
+  const model = buildOpenRouterModel(openRouterModel({
+    context_length: 1_048_576,
+    top_provider: {
+      context_length: 32_000,
+      max_completion_tokens: 8_192,
+    },
+  }), undefined);
+
+  expect(model).toMatchObject({
+    limit: {
+      context: 1_048_576,
+      output: 8_192,
+    },
+  });
+});
+
 test("factors OpenRouter Pro routes against canonical OpenAI metadata", () => {
   const model = buildOpenRouterModel(openRouterModel({
     id: "openai/gpt-5.6-sol-pro",
diff --git a/providers/openrouter/models/xiaomi/mimo-v2.5.toml b/providers/openrouter/models/xiaomi/mimo-v2.5.toml
index 741832ed73..4e5f84fe55 100644
--- a/providers/openrouter/models/xiaomi/mimo-v2.5.toml
+++ b/providers/openrouter/models/xiaomi/mimo-v2.5.toml
@@ -11,6 +11,3 @@ type = "toggle"
 input = 0.105
 output = 0.28
 cache_read = 0.028
-
-[limit]
-context = 32_000