fix: disable Tencent Hy3 free model

apps/web/src/app/api/openrouter/[...path]/route.test.ts

@@ -27,7 +27,7 @@ import {
   logFreeModelRequest,
 } from '@/lib/free-model-rate-limiter';
 import { gemma_4_26b_a4b_it_free_model } from '@/lib/ai-gateway/providers/google';
-import { tencent_hy3_free_model } from '@/lib/ai-gateway/providers/tencent';
+import { stepfun_37_flash_free_model } from '@/lib/ai-gateway/providers/stepfun';
 import { getEffectiveModelDecision } from '@/lib/organizations/effective-model-access.server';
 import { getPercentageRoutedPartnerProvider } from '@/lib/ai-gateway/providers/partner/routing';
 import { FRIENDLI_GLM_PUBLIC_ID } from '@/lib/ai-gateway/providers/partner/constants';
@@ -608,7 +608,9 @@ describe('POST /api/openrouter/v1/chat/completions rules-engine actions', () =>
     } as unknown as Awaited<ReturnType<typeof getUserFromAuth>>);
 
     const { POST } = await import('./route');
-    const response = await POST(makeRequest(makeBody(tencent_hy3_free_model.public_id)) as never);
+    const response = await POST(
+      makeRequest(makeBody(stepfun_37_flash_free_model.public_id)) as never
+    );
 
     expect(response.status).toBe(200);
     expect(mockedCheckFreeModelRateLimit).not.toHaveBeenCalled();

apps/web/src/lib/ai-gateway/models.test.ts

@@ -81,11 +81,13 @@ describe('isFreeModel', () => {
       expect(findKiloExclusiveModel('qwen/qwen3.7-plus')).toBeNull();
     });
 
-    test('registers Tencent Hy3 as an Auto Free model', () => {
-      expect(findKiloExclusiveModel('tencent/hy3:free')).toBe(tencent_hy3_free_model);
-      expect(tencent_hy3_free_model.internal_id).toBe('tencent/hy3');
-      expect(tencent_hy3_free_model.inference_provider_restriction).toEqual(['tencent']);
-      expect(autoFreeModels.map(({ model }) => model)).toContain(tencent_hy3_free_model.public_id);
+    test('disables Tencent Hy3 free and excludes it from Auto Free and preferred models', () => {
+      expect(tencent_hy3_free_model.status).toBe('disabled');
+      expect(findKiloExclusiveModel('tencent/hy3:free')).toBeNull();
+      expect(autoFreeModels.map(({ model }) => model)).not.toContain(
+        tencent_hy3_free_model.public_id
+      );
+      expect(preferredModels).not.toContain(tencent_hy3_free_model.public_id);
     });
 
     test('registers LongCat 2.0 as an Auto Free model', async () => {
@@ -231,18 +233,16 @@ describe('isFreeModel', () => {
         Object.fromEntries(autoFreeModels.map(({ model, reasoning }) => [model, reasoning]))
       ).toEqual({
         'stepfun/step-3.7-flash:free': { enabled: true, effort: 'high' },
-        'tencent/hy3:free': { enabled: true, effort: 'high' },
         'poolside/laguna-s-2.1:free': { enabled: true, effort: 'high' },
         'meituan/longcat-2.0-free': { enabled: true, effort: 'high' },
       });
     });
 
-    test('weights Auto Free models at 70% StepFun and 10% each for Hy3, Laguna, and LongCat', () => {
+    test('weights Auto Free models at 7:1:1 for StepFun, Laguna, and LongCat', () => {
       expect(
         Object.fromEntries(autoFreeModels.map(({ model, weight }) => [model, weight]))
       ).toEqual({
         'stepfun/step-3.7-flash:free': 7,
-        'tencent/hy3:free': 1,
         'poolside/laguna-s-2.1:free': 1,
         'meituan/longcat-2.0-free': 1,
       });
@@ -409,12 +409,12 @@ describe('getKiloExclusiveInferenceProviderRestriction', () => {
     expect(
       getKiloExclusiveInferenceProviderRestriction(gpt_5_6_sol_discounted_model.public_id)
     ).toEqual(new Set(['openai']));
-    expect(getKiloExclusiveInferenceProviderRestriction(tencent_hy3_free_model.public_id)).toEqual(
-      new Set(['tencent'])
-    );
   });
 
-  test('does not treat unrestricted exclusives or unknown ids as restricted', () => {
+  test('does not treat disabled or unrestricted exclusives or unknown ids as restricted', () => {
+    expect(
+      getKiloExclusiveInferenceProviderRestriction(tencent_hy3_free_model.public_id)
+    ).toBeUndefined();
     expect(
       getKiloExclusiveInferenceProviderRestriction(gemma_4_26b_a4b_it_free_model.public_id)
     ).toBeUndefined();

apps/web/src/lib/ai-gateway/providers/tencent.ts

@@ -7,7 +7,7 @@ export const tencent_hy3_free_model: KiloExclusiveModel = {
     'Hy3 is a 295B-parameter Mixture-of-Experts model from Tencent, activating 21B parameters per token. It supports configurable reasoning effort, agentic workflows, reliable tool calling, and long-context tasks across coding, document processing, financial analysis, and frontend development.',
   context_length: 262_144,
   max_completion_tokens: 128_000,
-  status: 'public',
+  status: 'disabled',
   flags: ['reasoning'],
   gateway: 'openrouter',
   internal_id: 'tencent/hy3',

apps/web/src/lib/ai-gateway/unavailable-models.test.ts

@@ -10,11 +10,15 @@ describe('unavailable models', () => {
     expect(isUnavailableModel('sakana/fugu-ultra')).toBe(false);
     expect(isUnavailableModel('google/gemma-4-31b-it:free')).toBe(true);
     expect(isUnavailableModel('openai/gpt-oss-20b')).toBe(false);
+    expect(isUnavailableModel('tencent/hy3:free')).toBe(true);
+    expect(isUnavailableModel('tencent/hy3')).toBe(false);
   });
 
   test('matches normalized families for provider metadata', () => {
     expect(familyHasUnavailableFreeModel('openai/gpt-oss-20b:free')).toBe(true);
     expect(familyHasUnavailableFreeModel('openai/gpt-oss-20b')).toBe(true);
     expect(familyHasUnavailableFreeModel('cohere/north-mini-code')).toBe(false);
+    expect(familyHasUnavailableFreeModel('tencent/hy3:free')).toBe(true);
+    expect(familyHasUnavailableFreeModel('tencent/hy3')).toBe(true);
   });
 });

apps/web/src/lib/ai-gateway/unavailable-models.ts

@@ -7,6 +7,7 @@ const unavailableModelIds: ReadonlySet<string> = new Set([
   'nvidia/nemotron-nano-12b-v2-vl:free',
   'nvidia/nemotron-nano-9b-v2:free',
   'openai/gpt-oss-20b:free',
+  'tencent/hy3:free',
   'z-ai/glm-5.2:free',
 ]);
 

apps/web/src/tests/openrouter-models-config.test.ts

@@ -16,7 +16,6 @@ describe('OpenRouter Models Config', () => {
       CLAUDE_OPUS_CURRENT_MODEL_ID,
       GPT_CURRENT_MODEL_ID,
       'z-ai/glm-5.3',
-      tencent_hy3_free_model.public_id,
     ];
 
     expectedModels.forEach(model => {
@@ -29,6 +28,7 @@ describe('OpenRouter Models Config', () => {
       'stealth/qwen3.6-plus',
       QWEN37_PLUS_MODEL_ID,
       'deepseek/deepseek-v4-pro',
+      tencent_hy3_free_model.public_id,
     ];
 
     supersededModels.forEach(model => {