From 8d8a3d942a45de94ed0f8023c4eccc0e6a9e7ed2 Mon Sep 17 00:00:00 2001
From: chrarnoldus <12196001+chrarnoldus@users.noreply.github.com>
Date: Fri, 11 Sep 2026 18:09:51 +0000
Subject: [PATCH] refactor(ai-gateway): remove Perplexity partner provider

Co-authored-by: kiloconnect[bot] <240665456+kiloconnect[bot]@users.noreply.github.com>
---
 ENVIRONMENT.md                                |   1 -
 .../src/app/admin/gateway/RoutingContent.tsx  |  48 +---
 .../api/openrouter/[...path]/route.test.ts    | 248 +-----------------
 .../src/app/api/openrouter/[...path]/route.ts |  78 +-----
 .../src/lib/ai-gateway/custom-pricing.test.ts |  68 ++---
 apps/web/src/lib/ai-gateway/custom-pricing.ts |  10 +-
 .../src/lib/ai-gateway/gateway-config.test.ts |  70 ++---
 apps/web/src/lib/ai-gateway/gateway-config.ts |   5 -
 .../apply-provider-specific-logic.test.ts     |  10 +-
 .../apply-provider-specific-logic.ts          |  20 +-
 .../ai-gateway/providers/partner/constants.ts |   2 -
 .../ai-gateway/providers/partner/pricing.ts   |  24 --
 .../providers/partner/providers.test.ts       |  80 ------
 .../ai-gateway/providers/partner/providers.ts |  34 ---
 .../providers/partner/routing.test.ts         | 159 -----------
 .../ai-gateway/providers/partner/routing.ts   |  74 ------
 .../providers/provider-definitions.test.ts    |   1 -
 .../ai-gateway/providers/routing-config.ts    |   4 -
 .../providers/routing-percentage.ts           |   4 +-
 .../web/src/lib/ai-gateway/providers/types.ts |   1 -
 .../ai-gateway/providers/vercel/index.test.ts |   1 -
 .../ai-gateway/rewriteModelResponse.test.ts   |  19 --
 .../lib/ai-gateway/rewriteModelResponse.ts    |   3 +-
 .../routers/admin/gateway-config-router.ts    |   1 -
 24 files changed, 68 insertions(+), 897 deletions(-)
 delete mode 100644 apps/web/src/lib/ai-gateway/providers/partner/constants.ts
 delete mode 100644 apps/web/src/lib/ai-gateway/providers/partner/pricing.ts
 delete mode 100644 apps/web/src/lib/ai-gateway/providers/partner/providers.test.ts
 delete mode 100644 apps/web/src/lib/ai-gateway/providers/partner/providers.ts
 delete mode 100644 apps/web/src/lib/ai-gateway/providers/partner/routing.test.ts
 delete mode 100644 apps/web/src/lib/ai-gateway/providers/partner/routing.ts

diff --git a/ENVIRONMENT.md b/ENVIRONMENT.md
index f4e6547e2e..b9d9e1cf66 100644
--- a/ENVIRONMENT.md
+++ b/ENVIRONMENT.md
@@ -269,7 +269,6 @@ Manage shared web env var additions and rotations with `pnpm web:env set <VARIAB
 - `MISTRAL_API_KEY` - Mistral API key; used in `apps/web/src/lib/ai-gateway/embeddings/embedding-providers.ts` for `codestral-embed-2505` and `mistral-embed` embeddings, in the FIM completions proxy at `apps/web/src/app/api/fim/completions/route.ts` (routes Mistral Codestral vs. La Plateforme keys), and as a provider config in `apps/web/src/lib/config.server.ts`. `[SECRET]`
 - `LONGCAT_API_KEY` - LongCat API key for model inference through the AI gateway; provider definition in `apps/web/src/lib/ai-gateway/providers/definitions/longcat.ts`. `[SECRET]`
 - `STREAMLAKE_API_KEY` - StreamLake API key for model inference through the AI gateway; provider definition in `apps/web/src/lib/ai-gateway/providers/definitions/streamlake.ts`. `[SECRET]`
-- `PERPLEXITY_API_KEY` - Perplexity API key for percentage-routed Kimi K3 inference through the AI gateway; provider definition in `apps/web/src/lib/ai-gateway/providers/partner/providers.ts`. `[SECRET]`
 - `INCEPTION_API_KEY` - Inception Labs API key; used in `apps/web/src/app/api/fim/completions/route.ts` and `apps/web/src/app/api/edit/completions/route.ts` as a fill-in-the-middle (FIM) provider, with endpoint `https://api.inceptionlabs.ai/v1/fim/completions`. Defined in `apps/web/src/lib/config.server.ts`. `[SECRET]`
 - `AI_ATTRIBUTION_ADMIN_SECRET` - Admin secret for the AI Attribution service (`apps/web/src/lib/ai-attribution-service.ts`); sent as `X-Admin-Secret` header. `[SECRET]`
 - `ARTIFICIAL_ANALYSIS_API_KEY` - API key for Artificial Analysis (`apps/web/src/lib/model-stats/sync-artificial-analysis.ts`); sent as `x-api-key` header for model benchmarking data sync. `[SECRET]`
diff --git a/apps/web/src/app/admin/gateway/RoutingContent.tsx b/apps/web/src/app/admin/gateway/RoutingContent.tsx
index eb5329b064..e0a3e3af67 100644
--- a/apps/web/src/app/admin/gateway/RoutingContent.tsx
+++ b/apps/web/src/app/admin/gateway/RoutingContent.tsx
@@ -10,7 +10,6 @@ import { Textarea } from '@/components/ui/textarea';
 import { Label } from '@/components/ui/label';
 import { Card, CardContent, CardDescription, CardHeader, CardTitle } from '@/components/ui/card';
 import {
-  DEFAULT_PERPLEXITY_PERCENTAGE,
   DEFAULT_VERCEL_PERCENTAGE,
   DEFAULT_VERCEL_PERCENTAGE_FREE,
   NOTE_MAX_LENGTH,
@@ -25,7 +24,6 @@ export function RoutingContent() {
 
   const [inputValue, setInputValue] = useState('');
   const [freeInputValue, setFreeInputValue] = useState('');
-  const [perplexityInputValue, setPerplexityInputValue] = useState('');
   const [optOutModelsValue, setOptOutModelsValue] = useState('');
   const [noteValue, setNoteValue] = useState('');
   const [hasChanges, setHasChanges] = useState(false);
@@ -34,7 +32,6 @@ export function RoutingContent() {
     if (data) {
       setInputValue(data.vercel_routing_percentage?.toString() ?? '');
       setFreeInputValue(data.vercel_routing_percentage_free?.toString() ?? '');
-      setPerplexityInputValue(data.perplexity_routing_percentage?.toString() ?? '');
       setOptOutModelsValue(data.vercel_routing_opt_out_models.join('\n'));
       setNoteValue(data.note ?? '');
       setHasChanges(false);
@@ -83,8 +80,7 @@ export function RoutingContent() {
     const note = noteInput();
     const paid = parsePercentage(inputValue);
     const free = parsePercentage(freeInputValue);
-    const perplexity = parsePercentage(perplexityInputValue);
-    if (paid === undefined || free === undefined || perplexity === undefined) {
+    if (paid === undefined || free === undefined) {
       toast.error(
         'Enter a percentage between 0 and 100 with up to 3 decimal places, or leave it empty for the default.'
       );
@@ -94,7 +90,6 @@ export function RoutingContent() {
       vercel_routing_percentage: paid,
       vercel_routing_percentage_free: free,
       vercel_routing_opt_out_models: optOutModels(),
-      perplexity_routing_percentage: perplexity,
       note,
     });
   }
@@ -104,7 +99,6 @@ export function RoutingContent() {
       vercel_routing_percentage: null,
       vercel_routing_percentage_free: null,
       vercel_routing_opt_out_models: optOutModels(),
-      perplexity_routing_percentage: null,
       note: noteInput(),
     });
   }
@@ -115,11 +109,9 @@ export function RoutingContent() {
 
   const currentOverride = data?.vercel_routing_percentage;
   const currentFreeOverride = data?.vercel_routing_percentage_free;
-  const currentPerplexityOverride = data?.perplexity_routing_percentage;
   const isOverrideActive =
     (currentOverride !== null && currentOverride !== undefined) ||
-    (currentFreeOverride !== null && currentFreeOverride !== undefined) ||
-    (currentPerplexityOverride !== null && currentPerplexityOverride !== undefined);
+    (currentFreeOverride !== null && currentFreeOverride !== undefined);
 
   return (
     <div className="flex w-full flex-col gap-y-6">
@@ -131,8 +123,7 @@ export function RoutingContent() {
             to Vercel (vs OpenRouter). Models not available on Vercel always go to OpenRouter, so
             overall traffic may still be skewed towards OpenRouter. Paid and free models are
             configured separately. Leave empty to use the default ({DEFAULT_VERCEL_PERCENTAGE}% for
-            paid, {DEFAULT_VERCEL_PERCENTAGE_FREE}% for free). Perplexity routes only its configured
-            models and defaults to {DEFAULT_PERPLEXITY_PERCENTAGE}%.
+            paid, {DEFAULT_VERCEL_PERCENTAGE_FREE}% for free).
           </CardDescription>
         </CardHeader>
         <CardContent className="flex flex-col gap-4">
@@ -183,30 +174,6 @@ export function RoutingContent() {
             />
             <span className="text-muted-foreground text-sm">%</span>
           </div>
-          <p className="pt-2 text-sm font-medium">Direct partners</p>
-          <div className="flex items-center gap-3">
-            <Label htmlFor="routing-perplexity" className="w-24 shrink-0">
-              Perplexity
-            </Label>
-            <Input
-              id="routing-perplexity"
-              type="number"
-              min={0}
-              max={100}
-              step={0.001}
-              placeholder={`Default: ${DEFAULT_PERPLEXITY_PERCENTAGE}%`}
-              value={perplexityInputValue}
-              onChange={e => {
-                setPerplexityInputValue(e.target.value);
-                setHasChanges(true);
-              }}
-              onKeyDown={e => {
-                if (e.key === 'Enter') handleSave();
-              }}
-              className="w-48"
-            />
-            <span className="text-muted-foreground text-sm">%</span>
-          </div>
           <div className="flex flex-col gap-2">
             <Label htmlFor="routing-opt-out-models">Opt-out models</Label>
             <Textarea
@@ -269,11 +236,7 @@ export function RoutingContent() {
                 <span className="text-foreground font-medium">
                   {currentFreeOverride ?? DEFAULT_VERCEL_PERCENTAGE_FREE}%
                 </span>{' '}
-                of free traffic goes to Vercel. Perplexity receives{' '}
-                <span className="text-foreground font-medium">
-                  {currentPerplexityOverride ?? DEFAULT_PERPLEXITY_PERCENTAGE}%
-                </span>{' '}
-                of eligible Kimi K3 traffic.
+                of free traffic goes to Vercel.
                 {data?.updated_by_email && (
                   <span className="ml-1">
                     Set by {data.updated_by_email}
@@ -284,8 +247,7 @@ export function RoutingContent() {
             ) : (
               <p>
                 No override set. Using default routing ({DEFAULT_VERCEL_PERCENTAGE}% of paid and{' '}
-                {DEFAULT_VERCEL_PERCENTAGE_FREE}% of free traffic to Vercel and{' '}
-                {DEFAULT_PERPLEXITY_PERCENTAGE}% to Perplexity).
+                {DEFAULT_VERCEL_PERCENTAGE_FREE}% of free traffic to Vercel).
               </p>
             )}
             {data?.note && (
diff --git a/apps/web/src/app/api/openrouter/[...path]/route.test.ts b/apps/web/src/app/api/openrouter/[...path]/route.test.ts
index 3d65029d8b..a34c148646 100644
--- a/apps/web/src/app/api/openrouter/[...path]/route.test.ts
+++ b/apps/web/src/app/api/openrouter/[...path]/route.test.ts
@@ -36,10 +36,6 @@ import {
 import { gemma_4_26b_a4b_it_free_model } from '@/lib/ai-gateway/providers/google';
 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 { PERPLEXITY_KIMI_PUBLIC_ID } from '@/lib/ai-gateway/providers/partner/constants';
-import type { GatewayMessagesRequest } from '@/lib/ai-gateway/providers/openrouter/types';
-import { warnExceptInTest } from '@/lib/utils.server';
 
 jest.mock('next/server', () => {
   return {
@@ -76,7 +72,6 @@ jest.mock('@/lib/ai-gateway/abuse-service', () => {
   };
 });
 jest.mock('@/lib/ai-gateway/providers/get-provider');
-jest.mock('@/lib/ai-gateway/providers/partner/routing');
 jest.mock('@/lib/ai-gateway/providers/direct-byok', () => ({
   getDirectByokModel: jest.fn(async () => ({ provider: null, model: null })),
 }));
@@ -111,10 +106,6 @@ jest.mock('@/lib/ai-gateway/llm-proxy-helpers', () => {
   };
 });
 jest.mock('@/lib/ai-gateway/auto-routing-decision');
-jest.mock('@/lib/utils.server', () => ({
-  ...jest.requireActual('@/lib/utils.server'),
-  warnExceptInTest: jest.fn(),
-}));
 jest.mock('@/lib/ai-gateway/auto-routing-denied-models', () => ({
   collectDeniedAutoRoutingModelIds: jest.fn().mockResolvedValue([]),
 }));
@@ -155,8 +146,6 @@ const mockedCheckFreeModelRateLimitByUser = jest.mocked(checkFreeModelRateLimitB
 const mockedCheckPromotionLimit = jest.mocked(checkPromotionLimit);
 const mockedLogFreeModelRequest = jest.mocked(logFreeModelRequest);
 const mockedGetEffectiveModelDecision = jest.mocked(getEffectiveModelDecision);
-const mockedGetPercentageRoutedPartnerProvider = jest.mocked(getPercentageRoutedPartnerProvider);
-const mockedWarnExceptInTest = jest.mocked(warnExceptInTest);
 
 const provider = {
   id: 'openrouter',
@@ -169,27 +158,6 @@ const provider = {
   transformRequest: jest.fn(),
 } satisfies Provider;
 
-const vercelProvider = {
-  ...provider,
-  id: 'vercel',
-  apiUrl: 'https://ai-gateway.vercel.sh/v1',
-  transformRequest: jest.fn(),
-} satisfies Provider;
-
-const partnerProvider = {
-  ...provider,
-  id: 'perplexity',
-  apiUrl: 'https://api.perplexity.ai/router/v1',
-  supportedChatApis: ['messages'],
-  transformRequest: jest.fn(async context => {
-    Object.assign(context.request.body, {
-      model: 'perplexity/kimi-k3',
-      metadata: { transformedBy: 'perplexity' },
-    });
-    delete context.request.body.provider;
-  }),
-} satisfies Provider;
-
 function makeRequest(body: unknown, headers?: HeadersInit) {
   return new Request('http://localhost:3000/api/openrouter/v1/chat/completions', {
     method: 'POST',
@@ -202,17 +170,6 @@ function makeRequest(body: unknown, headers?: HeadersInit) {
   });
 }
 
-function makeMessagesRequest(body: unknown) {
-  return new Request('http://localhost:3000/api/openrouter/v1/messages', {
-    method: 'POST',
-    headers: {
-      'Content-Type': 'application/json',
-      'x-forwarded-for': '127.0.0.1',
-    },
-    body: JSON.stringify(body),
-  });
-}
-
 function makeBody(model = 'openai/gpt-4o') {
   return {
     model,
@@ -714,7 +671,7 @@ describe('POST /api/openrouter/v1/chat/completions rules-engine actions', () =>
     });
     mockedGetProvider.mockResolvedValue({
       kind: 'provider',
-      provider: { ...provider, id: 'perplexity' },
+      provider: { ...provider, id: 'martian' },
       userByok: null,
       bypassAccessCheck: false,
       experiment: {
@@ -749,7 +706,7 @@ describe('POST /api/openrouter/v1/chat/completions rules-engine actions', () =>
     });
     mockedGetProvider.mockResolvedValue({
       kind: 'provider',
-      provider: { ...provider, id: 'perplexity' },
+      provider: { ...provider, id: 'martian' },
       userByok: null,
       bypassAccessCheck: false,
       experiment: {
@@ -1513,204 +1470,3 @@ describe('auto-routing shadow classifier', () => {
     expect(mockedAfter).not.toHaveBeenCalled();
   });
 });
-
-describe('percentage-routed partner fallback', () => {
-  beforeEach(() => {
-    jest.clearAllMocks();
-    setUserAuth();
-    mockedGetProvider.mockResolvedValue({
-      kind: 'provider',
-      provider,
-      userByok: null,
-      bypassAccessCheck: false,
-    });
-    mockedGetPercentageRoutedPartnerProvider.mockResolvedValue(partnerProvider);
-    mockedClassifyAbuse.mockResolvedValue(classifyResult(null));
-    mockedRedisGet.mockResolvedValue(null);
-    mockedRedisSet.mockResolvedValue('OK');
-    mockedGetOpenRouterModels.mockResolvedValue(new Set());
-    mockedIsValidOpenRouterModelId.mockResolvedValue(true);
-    mockedEmitApiMetricsForResponse.mockReturnValue(undefined);
-    mockedAccountForMicrodollarUsage.mockReturnValue(undefined);
-  });
-
-  it.each([
-    ['openrouter', provider, 400],
-    ['vercel', vercelProvider, 500],
-  ] as const)(
-    'retries a failed partner request through the initial %s provider',
-    async (_sourceId, sourceProvider, partnerStatus) => {
-      mockedGetProvider.mockResolvedValue({
-        kind: 'provider',
-        provider: sourceProvider,
-        userByok: null,
-        bypassAccessCheck: false,
-      });
-      const failedPartnerResponse = new Response('partner failed', { status: partnerStatus });
-      jest.spyOn(failedPartnerResponse, 'clone').mockReturnValue(new Response('partner failed'));
-      const cancelPartnerBody = jest.spyOn(failedPartnerResponse.body!, 'cancel');
-      mockedUpstreamRequest
-        .mockResolvedValueOnce({ type: 'success', response: failedPartnerResponse })
-        .mockResolvedValueOnce({
-          type: 'success',
-          response: upstreamJsonResponse({ type: 'message', id: 'fallback-response' }),
-        });
-
-      const { POST } = await import('./route');
-      const response = await POST(
-        makeMessagesRequest({
-          model: PERPLEXITY_KIMI_PUBLIC_ID,
-          max_tokens: 1_024,
-          messages: [{ role: 'user', content: 'hello' }],
-        }) as never
-      );
-
-      expect(response.status).toBe(200);
-      expect(mockedUpstreamRequest).toHaveBeenCalledTimes(2);
-      const partnerAttempt = mockedUpstreamRequest.mock.calls[0]?.[0];
-      const fallbackAttempt = mockedUpstreamRequest.mock.calls[1]?.[0];
-      expect(partnerAttempt.provider).toBe(partnerProvider);
-      expect(partnerAttempt.body.model).toBe('perplexity/kimi-k3');
-      expect(partnerAttempt.body.metadata).toEqual({ transformedBy: 'perplexity' });
-      expect(fallbackAttempt.provider).toBe(sourceProvider);
-      expect(fallbackAttempt.body.model).toBe(PERPLEXITY_KIMI_PUBLIC_ID);
-      expect(fallbackAttempt.body.metadata).not.toHaveProperty('transformedBy');
-      expect(fallbackAttempt.body).not.toBe(partnerAttempt.body);
-      expect((fallbackAttempt.body as GatewayMessagesRequest).messages).not.toBe(
-        (partnerAttempt.body as GatewayMessagesRequest).messages
-      );
-      expect(fallbackAttempt.extraHeaders).not.toBe(partnerAttempt.extraHeaders);
-      expect(cancelPartnerBody).toHaveBeenCalledTimes(1);
-      const { after: mockedAfter } = jest.requireMock<{ after: jest.Mock }>('next/server');
-      const partnerLog = mockedAfter.mock.calls[0]?.[0];
-      expect(partnerLog).toBeInstanceOf(Promise);
-      await partnerLog;
-      expect(mockedWarnExceptInTest).toHaveBeenCalledWith(
-        'Partner request failed before managed fallback',
-        {
-          partner_provider: partnerProvider.id,
-          fallback_provider: sourceProvider.id,
-          status_code: partnerStatus,
-          body: 'partner failed',
-        }
-      );
-      expect(mockedEmitApiMetricsForResponse).toHaveBeenCalledWith(
-        expect.objectContaining({ provider: sourceProvider.id, statusCode: 200 }),
-        expect.any(Response),
-        expect.any(Number)
-      );
-      expect(mockedAccountForMicrodollarUsage.mock.calls[0]?.[1]).toMatchObject({
-        provider: sourceProvider.id,
-        status_code: 200,
-      });
-      expect(mockedRewriteModelResponse).toHaveBeenCalledWith(
-        expect.objectContaining({ providerId: sourceProvider.id })
-      );
-    }
-  );
-
-  it('logs partner response body read errors before falling back', async () => {
-    const failedPartnerResponse = upstreamJsonResponse({ error: 'partner failed' }, 500);
-    const unreadableResponse = new Response(
-      new ReadableStream({
-        start(controller) {
-          controller.error(new Error('read failed'));
-        },
-      })
-    );
-    jest.spyOn(failedPartnerResponse, 'clone').mockReturnValue(unreadableResponse);
-    mockedUpstreamRequest
-      .mockResolvedValueOnce({ type: 'success', response: failedPartnerResponse })
-      .mockResolvedValueOnce({
-        type: 'success',
-        response: upstreamJsonResponse({ type: 'message', id: 'fallback-response' }),
-      });
-
-    const { POST } = await import('./route');
-    const response = await POST(
-      makeMessagesRequest({
-        model: PERPLEXITY_KIMI_PUBLIC_ID,
-        max_tokens: 1_024,
-        messages: [{ role: 'user', content: 'hello' }],
-      }) as never
-    );
-    const { after: mockedAfter } = jest.requireMock<{ after: jest.Mock }>('next/server');
-    const partnerLog = mockedAfter.mock.calls[0]?.[0];
-    expect(partnerLog).toBeInstanceOf(Promise);
-    await partnerLog;
-
-    expect(response.status).toBe(200);
-    expect(mockedWarnExceptInTest).toHaveBeenCalledWith(
-      'Partner request failed before managed fallback',
-      {
-        partner_provider: partnerProvider.id,
-        fallback_provider: provider.id,
-        status_code: 500,
-        response_body_read_error: 'Error: read failed',
-      }
-    );
-  });
-
-  it('does not retry a successful partner response', async () => {
-    mockedUpstreamRequest.mockResolvedValue({
-      type: 'success',
-      response: upstreamJsonResponse({ type: 'message', id: 'partner-response' }),
-    });
-
-    const { POST } = await import('./route');
-    const response = await POST(
-      makeMessagesRequest({
-        model: PERPLEXITY_KIMI_PUBLIC_ID,
-        max_tokens: 1_024,
-        messages: [{ role: 'user', content: 'hello' }],
-      }) as never
-    );
-
-    expect(response.status).toBe(200);
-    expect(mockedUpstreamRequest).toHaveBeenCalledTimes(1);
-    expect(mockedUpstreamRequest.mock.calls[0]?.[0].provider).toBe(partnerProvider);
-  });
-
-  it('returns a failed managed fallback without a third attempt', async () => {
-    mockedUpstreamRequest
-      .mockResolvedValueOnce({
-        type: 'success',
-        response: upstreamJsonResponse({ error: 'partner failed' }, 500),
-      })
-      .mockResolvedValueOnce({
-        type: 'success',
-        response: upstreamJsonResponse({ error: 'fallback failed' }, 503),
-      });
-
-    const { POST } = await import('./route');
-    const response = await POST(
-      makeMessagesRequest({
-        model: PERPLEXITY_KIMI_PUBLIC_ID,
-        max_tokens: 1_024,
-        messages: [{ role: 'user', content: 'hello' }],
-      }) as never
-    );
-
-    expect(response.status).toBe(503);
-    expect(mockedUpstreamRequest).toHaveBeenCalledTimes(2);
-  });
-
-  it('does not retry a partner transport failure without an HTTP status', async () => {
-    mockedUpstreamRequest.mockResolvedValue({
-      type: 'error',
-      response: new Response(null, { status: 503 }) as never,
-    });
-
-    const { POST } = await import('./route');
-    const response = await POST(
-      makeMessagesRequest({
-        model: PERPLEXITY_KIMI_PUBLIC_ID,
-        max_tokens: 1_024,
-        messages: [{ role: 'user', content: 'hello' }],
-      }) as never
-    );
-
-    expect(response.status).toBe(503);
-    expect(mockedUpstreamRequest).toHaveBeenCalledTimes(1);
-  });
-});
diff --git a/apps/web/src/app/api/openrouter/[...path]/route.ts b/apps/web/src/app/api/openrouter/[...path]/route.ts
index 6ed31fb28f..fe278f409a 100644
--- a/apps/web/src/app/api/openrouter/[...path]/route.ts
+++ b/apps/web/src/app/api/openrouter/[...path]/route.ts
@@ -15,10 +15,7 @@ import type {
   GatewayMessagesRequest,
   GatewayRequest,
 } from '@/lib/ai-gateway/providers/openrouter/types';
-import {
-  getProvider,
-  type GetProviderProviderResult,
-} from '@/lib/ai-gateway/providers/get-provider';
+import { getProvider } from '@/lib/ai-gateway/providers/get-provider';
 import { getDirectByokModel } from '@/lib/ai-gateway/providers/direct-byok';
 import { sendUpstreamAttempt } from '@/lib/ai-gateway/providers/upstream-attempt';
 import { debugSaveProxyRequest } from '@/lib/debugUtils';
@@ -65,7 +62,6 @@ import {
   rewriteModelResponse,
   logUnrewrittenResponse,
 } from '@/lib/ai-gateway/rewriteModelResponse';
-import { getPercentageRoutedPartnerProvider } from '@/lib/ai-gateway/providers/partner/routing';
 import {
   createAnonymousContext,
   isAnonymousContext,
@@ -112,7 +108,7 @@ import {
   hasMiddleOutTransform,
 } from '@/lib/ai-gateway/providers/openrouter/request-helpers';
 import { redactProviderHints } from '@kilocode/auto-routing-contracts';
-import { logExceptInTest, warnExceptInTest } from '@/lib/utils.server';
+import { logExceptInTest } from '@/lib/utils.server';
 import { readDb } from '@/lib/drizzle';
 import { getOrganizationGroupPolicyContext } from '@/lib/organizations/organization-group-policy-context.server';
 import {
@@ -899,29 +895,6 @@ export async function POST(request: NextRequest): Promise<NextResponseType<unkno
     }
   }
 
-  const partnerProvider = await getPercentageRoutedPartnerProvider({
-    requestedModel: effectiveModelIdLowerCased,
-    request: requestBodyParsed,
-    randomSeed: taskId || user.id,
-    sourceProviderId: effectiveProviderContext.provider.id,
-    hasUserByok: effectiveProviderContext.userByok !== null,
-  });
-  let partnerFallback:
-    | { providerContext: GetProviderProviderResult; request: GatewayRequest }
-    | undefined;
-  if (partnerProvider) {
-    partnerFallback = {
-      providerContext: effectiveProviderContext,
-      request: structuredClone(requestBodyParsed),
-    };
-    effectiveProviderContext = {
-      kind: 'provider',
-      provider: partnerProvider,
-      userByok: null,
-      bypassAccessCheck: false,
-    };
-  }
-
   console.debug(`Routing request to ${effectiveProviderContext.provider.id}`);
 
   // Extract properties for usage context after final provider selection.
@@ -1003,7 +976,7 @@ export async function POST(request: NextRequest): Promise<NextResponseType<unkno
     signal: request.signal,
     vercelRequestId,
   };
-  let attempt = await sendUpstreamAttempt({
+  const attempt = await sendUpstreamAttempt({
     ...upstreamAttemptOptions,
     providerContext: effectiveProviderContext,
     request: requestBodyParsed,
@@ -1014,51 +987,6 @@ export async function POST(request: NextRequest): Promise<NextResponseType<unkno
   }
   if (attempt.type === 'error') return attempt.response;
 
-  if (partnerFallback && attempt.response.status >= 400) {
-    const partnerFailureLog = {
-      partner_provider: effectiveProviderContext.provider.id,
-      fallback_provider: partnerFallback.providerContext.provider.id,
-      status_code: attempt.response.status,
-    };
-    const responseForLogging = attempt.response.clone();
-    after(
-      (async () => {
-        try {
-          warnExceptInTest('Partner request failed before managed fallback', {
-            ...partnerFailureLog,
-            body: await responseForLogging.text(),
-          });
-        } catch (error) {
-          warnExceptInTest('Partner request failed before managed fallback', {
-            ...partnerFailureLog,
-            response_body_read_error: String(error),
-          });
-        }
-      })()
-    );
-    try {
-      await attempt.response.body?.cancel();
-    } catch {
-      warnExceptInTest('Failed to cancel discarded partner response body');
-    }
-
-    effectiveProviderContext = partnerFallback.providerContext;
-    requestBodyParsed = partnerFallback.request;
-    usageContext.provider = effectiveProviderContext.provider.id;
-    usageContext.user_byok = !!effectiveProviderContext.userByok;
-
-    attempt = await sendUpstreamAttempt({
-      ...upstreamAttemptOptions,
-      providerContext: effectiveProviderContext,
-      request: requestBodyParsed,
-      delayMs: 0,
-    });
-    if (attempt.type === 'invalid-openrouter-model') {
-      return modelDoesNotExistOnOpenRouterResponse(effectiveModelIdLowerCased);
-    }
-    if (attempt.type === 'error') return attempt.response;
-  }
-
   const { response, toolsAvailable, toolsUsed, experimentPromptCapture } = attempt;
   if (experimentPromptCapture) usageContext.experimentPromptCapture = experimentPromptCapture;
   const finalUpstreamModel = requestBodyParsed.body.model ?? effectiveModelIdLowerCased;
diff --git a/apps/web/src/lib/ai-gateway/custom-pricing.test.ts b/apps/web/src/lib/ai-gateway/custom-pricing.test.ts
index 311e3f4d0f..b0221c4847 100644
--- a/apps/web/src/lib/ai-gateway/custom-pricing.test.ts
+++ b/apps/web/src/lib/ai-gateway/custom-pricing.test.ts
@@ -1,7 +1,6 @@
 import { describe, expect, test } from '@jest/globals';
 import { captureMessage } from '@sentry/nextjs';
 import type { OpenRouterModel } from '@/lib/organizations/organization-types';
-import { PERPLEXITY_KIMI_PUBLIC_ID } from '@/lib/ai-gateway/providers/partner/constants';
 import { GEMINI_FLASH_CURRENT_MODEL_ID } from '@/lib/ai-gateway/providers/google';
 import {
   applyCustomPricingToPricing,
@@ -118,56 +117,23 @@ describe('custom model pricing', () => {
     ).toBe(Math.round(262_144 * 0.96));
   });
 
-  test('uses Kimi K3 custom pricing only when market cost is missing', () => {
-    const usage = makeUsage({
-      inputTokens: 100,
-      outputTokens: 10,
-      cacheHitTokens: 20,
-      cacheWriteTokens: 30,
-    });
-
-    expect(calculateCustomCost_mUsd(PERPLEXITY_KIMI_PUBLIC_ID, usage)).toBe(
-      Math.round(50 * 3 + 10 * 15 + 20 * 0.3 + 30 * 3)
-    );
-    expect(
-      calculateCustomCost_mUsd(PERPLEXITY_KIMI_PUBLIC_ID, { ...usage, cost_mUsd: 123 })
-    ).toBeUndefined();
-  });
-
-  test('does not apply custom pricing to GLM 5.2', () => {
-    const model = makeModel('z-ai/glm-5.2');
-    const usage = makeUsage({
-      inputTokens: 100,
-      outputTokens: 10,
-      cacheHitTokens: 20,
-      cacheWriteTokens: 30,
-    });
-
-    expect(applyCustomPricingToModel(model)).toBe(model);
-    expect(applyCustomPricingToPricing(model.id, model.pricing)).toBe(model.pricing);
-    expect(calculateCustomCost_mUsd(model.id, usage)).toBeUndefined();
-    expect(calculateCustomCost_mUsd(model.id, { ...usage, cost_mUsd: 123 })).toBeUndefined();
-  });
-
-  test('uses fallback-only custom pricing in the model list', () => {
-    const model = makeModel(PERPLEXITY_KIMI_PUBLIC_ID);
-
-    expect(applyCustomPricingToModel(model)).toEqual({
-      ...model,
-      pricing: {
-        prompt: '0.000003000000',
-        completion: '0.000015000000',
-        input_cache_read: '0.000000300000',
-        input_cache_write: undefined,
-      },
-    });
-  });
-
-  test('preserves endpoint pricing for fallback-only custom pricing', () => {
-    const model = makeModel(PERPLEXITY_KIMI_PUBLIC_ID);
-
-    expect(applyCustomPricingToPricing(model.id, model.pricing)).toBe(model.pricing);
-  });
+  test.each(['z-ai/glm-5.2', 'moonshotai/kimi-k3'])(
+    'does not apply custom pricing to %s',
+    modelId => {
+      const model = makeModel(modelId);
+      const usage = makeUsage({
+        inputTokens: 100,
+        outputTokens: 10,
+        cacheHitTokens: 20,
+        cacheWriteTokens: 30,
+      });
+
+      expect(applyCustomPricingToModel(model)).toBe(model);
+      expect(applyCustomPricingToPricing(model.id, model.pricing)).toBe(model.pricing);
+      expect(calculateCustomCost_mUsd(model.id, usage)).toBeUndefined();
+      expect(calculateCustomCost_mUsd(model.id, { ...usage, cost_mUsd: 123 })).toBeUndefined();
+    }
+  );
 
   test('reports invalid negative uncached token counts', () => {
     const captureMessageMock = jest.mocked(captureMessage);
diff --git a/apps/web/src/lib/ai-gateway/custom-pricing.ts b/apps/web/src/lib/ai-gateway/custom-pricing.ts
index 5ac184ec13..3c269b85e4 100644
--- a/apps/web/src/lib/ai-gateway/custom-pricing.ts
+++ b/apps/web/src/lib/ai-gateway/custom-pricing.ts
@@ -3,7 +3,6 @@ import type { OpenRouterModel } from '@/lib/organizations/organization-types';
 import type { JustTheCostsUsageStats } from '@/lib/ai-gateway/processUsage.types';
 import { GEMINI_FLASH_CURRENT_MODEL_ID } from '@/lib/ai-gateway/providers/google';
 import { QWEN37_MAX_MODEL_ID, QWEN37_PLUS_MODEL_ID } from '@/lib/ai-gateway/providers/qwen';
-import { partnerPricingByModelId } from '@/lib/ai-gateway/providers/partner/pricing';
 import {
   calculateCost_mUsd,
   type Pricing,
@@ -19,8 +18,6 @@ export type CustomPricing = {
   pricing: PricingTiers;
   /** Human-readable discount shown in the model name; never used in calculations. */
   discountPercentage?: number;
-  /** Only use this pricing when the upstream response does not report a market cost. */
-  fallbackOnly?: boolean;
 };
 
 export const customPricingByModelId: Record<string, CustomPricing> = {
@@ -76,7 +73,6 @@ export const customPricingByModelId: Record<string, CustomPricing> = {
       },
     ],
   },
-  ...partnerPricingByModelId,
 };
 
 export function getCustomPricing(modelId: string): CustomPricing | undefined {
@@ -109,9 +105,7 @@ export function applyCustomPricingToPricing(
   pricing: OpenRouterModel['pricing']
 ): OpenRouterModel['pricing'] {
   const customPricing = getCustomPricing(modelId);
-  return customPricing && !customPricing.fallbackOnly
-    ? applyPricing(pricing, customPricing.pricing[0].pricing)
-    : pricing;
+  return customPricing ? applyPricing(pricing, customPricing.pricing[0].pricing) : pricing;
 }
 
 export function applyCustomPricingToModel(model: OpenRouterModel): OpenRouterModel {
@@ -135,7 +129,7 @@ export function calculateCustomCost_mUsd(
   usage: JustTheCostsUsageStats
 ): number | undefined {
   const customPricing = getCustomPricing(modelId);
-  if (!customPricing || (customPricing.fallbackOnly && usage.cost_mUsd > 0)) return undefined;
+  if (!customPricing) return undefined;
 
   const uncachedInputTokens = usage.inputTokens - usage.cacheHitTokens - usage.cacheWriteTokens;
   if (uncachedInputTokens < 0) {
diff --git a/apps/web/src/lib/ai-gateway/gateway-config.test.ts b/apps/web/src/lib/ai-gateway/gateway-config.test.ts
index 3a4691dae8..1c01582c29 100644
--- a/apps/web/src/lib/ai-gateway/gateway-config.test.ts
+++ b/apps/web/src/lib/ai-gateway/gateway-config.test.ts
@@ -23,7 +23,6 @@ describe('GatewayRoutingConfigSchema', () => {
       vercel_routing_percentage: 25,
       vercel_routing_percentage_free: null,
       vercel_routing_opt_out_models: [],
-      perplexity_routing_percentage: null,
     });
   });
 
@@ -32,7 +31,6 @@ describe('GatewayRoutingConfigSchema', () => {
       vercel_routing_percentage: null,
       vercel_routing_percentage_free: null,
       vercel_routing_opt_out_models: [],
-      perplexity_routing_percentage: null,
     });
   });
 
@@ -41,7 +39,6 @@ describe('GatewayRoutingConfigSchema', () => {
       vercel_routing_percentage: 25,
       vercel_routing_percentage_free: null,
       vercel_routing_opt_out_models: [],
-      perplexity_routing_percentage: null,
     });
   });
 
@@ -55,7 +52,6 @@ describe('GatewayRoutingConfigSchema', () => {
       vercel_routing_percentage: 25,
       vercel_routing_percentage_free: 80,
       vercel_routing_opt_out_models: [],
-      perplexity_routing_percentage: null,
     });
   });
 
@@ -69,21 +65,6 @@ describe('GatewayRoutingConfigSchema', () => {
       vercel_routing_percentage: 25,
       vercel_routing_percentage_free: null,
       vercel_routing_opt_out_models: ['moonshotai/kimi-k3'],
-      perplexity_routing_percentage: null,
-    });
-  });
-
-  test('accepts a Perplexity percentage', () => {
-    expect(
-      GatewayRoutingConfigSchema.parse({
-        vercel_routing_percentage: 25,
-        perplexity_routing_percentage: 20,
-      })
-    ).toEqual({
-      vercel_routing_percentage: 25,
-      vercel_routing_percentage_free: null,
-      vercel_routing_opt_out_models: [],
-      perplexity_routing_percentage: 20,
     });
   });
 
@@ -109,7 +90,6 @@ describe('GatewayConfigSchema', () => {
     });
     expect(parsed.note).toBeNull();
     expect(parsed.vercel_routing_opt_out_models).toEqual([]);
-    expect(parsed.perplexity_routing_percentage).toBeNull();
   });
 
   test('round-trips a note', () => {
@@ -128,32 +108,29 @@ describe('saved gateway configuration compatibility', () => {
   test.each([
     { name: 'runtime routing', schema: GatewayRoutingConfigSchema },
     { name: 'admin configuration', schema: GatewayConfigSchema },
-  ])(
-    '$name preserves Perplexity when saved JSON includes an obsolete Friendli percentage',
-    ({ schema }) => {
-      const savedJson = JSON.stringify({
-        vercel_routing_percentage: 25,
-        vercel_routing_percentage_free: 80,
-        vercel_routing_opt_out_models: ['moonshotai/kimi-k3'],
-        friendli_routing_percentage: 10,
-        perplexity_routing_percentage: 12.345,
-        updated_at: '2026-01-01T00:00:00.000Z',
-        updated_by: 'u1',
-        updated_by_email: 'a@example.com',
-        note: 'Existing routing overrides',
-      });
+  ])('$name drops obsolete partner percentages from saved JSON', ({ schema }) => {
+    const savedJson = JSON.stringify({
+      vercel_routing_percentage: 25,
+      vercel_routing_percentage_free: 80,
+      vercel_routing_opt_out_models: ['moonshotai/kimi-k3'],
+      friendli_routing_percentage: 10,
+      perplexity_routing_percentage: 12.345,
+      updated_at: '2026-01-01T00:00:00.000Z',
+      updated_by: 'u1',
+      updated_by_email: 'a@example.com',
+      note: 'Existing routing overrides',
+    });
 
-      const parsed = schema.parse(JSON.parse(savedJson));
+    const parsed = schema.parse(JSON.parse(savedJson));
 
-      expect(parsed).toMatchObject({
-        vercel_routing_percentage: 25,
-        vercel_routing_percentage_free: 80,
-        vercel_routing_opt_out_models: ['moonshotai/kimi-k3'],
-        perplexity_routing_percentage: 12.345,
-      });
-      expect(parsed).not.toHaveProperty('friendli_routing_percentage');
-    }
-  );
+    expect(parsed).toMatchObject({
+      vercel_routing_percentage: 25,
+      vercel_routing_percentage_free: 80,
+      vercel_routing_opt_out_models: ['moonshotai/kimi-k3'],
+    });
+    expect(parsed).not.toHaveProperty('friendli_routing_percentage');
+    expect(parsed).not.toHaveProperty('perplexity_routing_percentage');
+  });
 });
 
 describe('GatewayConfigInputSchema', () => {
@@ -163,14 +140,12 @@ describe('GatewayConfigInputSchema', () => {
         vercel_routing_percentage: 75,
         vercel_routing_percentage_free: 60,
         vercel_routing_opt_out_models: ['moonshotai/kimi-k3'],
-        perplexity_routing_percentage: 20,
         note: 'Rollout stable',
       })
     ).toEqual({
       vercel_routing_percentage: 75,
       vercel_routing_percentage_free: 60,
       vercel_routing_opt_out_models: ['moonshotai/kimi-k3'],
-      perplexity_routing_percentage: 20,
       note: 'Rollout stable',
     });
   });
@@ -181,14 +156,12 @@ describe('GatewayConfigInputSchema', () => {
         vercel_routing_percentage: null,
         vercel_routing_percentage_free: null,
         vercel_routing_opt_out_models: [],
-        perplexity_routing_percentage: null,
         note: null,
       })
     ).toEqual({
       vercel_routing_percentage: null,
       vercel_routing_percentage_free: null,
       vercel_routing_opt_out_models: [],
-      perplexity_routing_percentage: null,
       note: null,
     });
   });
@@ -199,7 +172,6 @@ describe('GatewayConfigInputSchema', () => {
         vercel_routing_percentage: 50,
         vercel_routing_percentage_free: 50,
         vercel_routing_opt_out_models: [],
-        perplexity_routing_percentage: 0,
         note: 'x'.repeat(NOTE_MAX_LENGTH + 1),
       })
     ).toThrow();
diff --git a/apps/web/src/lib/ai-gateway/gateway-config.ts b/apps/web/src/lib/ai-gateway/gateway-config.ts
index b427181bf9..83c0be19ce 100644
--- a/apps/web/src/lib/ai-gateway/gateway-config.ts
+++ b/apps/web/src/lib/ai-gateway/gateway-config.ts
@@ -2,7 +2,6 @@ import * as z from 'zod';
 
 export const DEFAULT_VERCEL_PERCENTAGE = 50;
 export const DEFAULT_VERCEL_PERCENTAGE_FREE = 50;
-export const DEFAULT_PERPLEXITY_PERCENTAGE = 0;
 
 export const RoutingPercentageSchema = z.number().min(0).max(100).multipleOf(0.001);
 
@@ -14,7 +13,6 @@ export const GatewayConfigSchema = z.object({
   vercel_routing_percentage: RoutingPercentageSchema.nullable(),
   vercel_routing_percentage_free: RoutingPercentageSchema.nullable().default(null),
   vercel_routing_opt_out_models: z.array(z.string().min(1)).default([]),
-  perplexity_routing_percentage: RoutingPercentageSchema.nullable().default(null),
   updated_at: z.string().nullable(),
   updated_by: z.string().nullable(),
   updated_by_email: z.string().nullable(),
@@ -27,7 +25,6 @@ export const DEFAULT_GATEWAY_CONFIG: GatewayConfig = {
   vercel_routing_percentage: null,
   vercel_routing_percentage_free: null,
   vercel_routing_opt_out_models: [],
-  perplexity_routing_percentage: null,
   updated_at: null,
   updated_by: null,
   updated_by_email: null,
@@ -49,7 +46,6 @@ export const GatewayRoutingConfigSchema = z.object({
   vercel_routing_percentage: RoutingPercentageSchema.nullable(),
   vercel_routing_percentage_free: RoutingPercentageSchema.nullable().default(null),
   vercel_routing_opt_out_models: z.array(z.string().min(1)).default([]),
-  perplexity_routing_percentage: RoutingPercentageSchema.nullable().default(null),
 });
 
 /** Schema for the admin set-mutation input. */
@@ -57,6 +53,5 @@ export const GatewayConfigInputSchema = z.object({
   vercel_routing_percentage: RoutingPercentageSchema.nullable(),
   vercel_routing_percentage_free: RoutingPercentageSchema.nullable(),
   vercel_routing_opt_out_models: z.array(z.string()),
-  perplexity_routing_percentage: RoutingPercentageSchema.nullable(),
   note: note.nullable(),
 });
diff --git a/apps/web/src/lib/ai-gateway/providers/apply-provider-specific-logic.test.ts b/apps/web/src/lib/ai-gateway/providers/apply-provider-specific-logic.test.ts
index c80e64120e..0b686330b5 100644
--- a/apps/web/src/lib/ai-gateway/providers/apply-provider-specific-logic.test.ts
+++ b/apps/web/src/lib/ai-gateway/providers/apply-provider-specific-logic.test.ts
@@ -14,8 +14,8 @@ import {
   type Provider,
   type ProviderId,
 } from '@/lib/ai-gateway/providers/types';
-import { PERPLEXITY_KIMI_PUBLIC_ID } from '@/lib/ai-gateway/providers/partner/constants';
 import { QWEN37_MAX_MODEL_ID } from '@/lib/ai-gateway/custom-pricing';
+import { KIMI_CURRENT_MODEL_ID } from '@/lib/ai-gateway/providers/moonshotai';
 import {
   gpt_5_6_sol_discounted_model,
   gpt_6_astra_flex_model,
@@ -38,7 +38,7 @@ function makeRequest(
 
 function makeProvider(responseTransforms: Provider['responseTransforms']): Provider {
   return {
-    id: 'perplexity',
+    id: 'openrouter',
     apiUrl: 'https://example.com/v1',
     apiUrlOverrides: {},
     apiKey: 'test-key',
@@ -67,7 +67,7 @@ function makeMessagesRequest(
 }
 
 describe('applyAnthropicThinkingDefault', () => {
-  it.each(['z-ai/glm-5.2', PERPLEXITY_KIMI_PUBLIC_ID, 'minimax/minimax-m3'])(
+  it.each(['z-ai/glm-5.2', KIMI_CURRENT_MODEL_ID, 'minimax/minimax-m3'])(
     'disables implicit thinking for %s',
     model => {
       const request = makeMessagesRequest(model);
@@ -114,7 +114,7 @@ describe('removeUnsupportedRequestServiceTier', () => {
     {
       model: QWEN37_MAX_MODEL_ID,
       kiloExclusiveModel: null,
-      reason: 'non-fallback custom pricing',
+      reason: 'custom pricing',
     },
     {
       model: gpt_5_6_sol_discounted_model.public_id,
@@ -145,7 +145,7 @@ describe('removeUnsupportedRequestServiceTier', () => {
   );
 
   it.each([
-    [PERPLEXITY_KIMI_PUBLIC_ID, null],
+    [KIMI_CURRENT_MODEL_ID, null],
     [gpt_6_astra_flex_model.public_id, gpt_6_astra_flex_model],
     ['vendor/standard-model', null],
   ] as const)('preserves the request-level tier for %s', (model, kiloExclusiveModel) => {
diff --git a/apps/web/src/lib/ai-gateway/providers/apply-provider-specific-logic.ts b/apps/web/src/lib/ai-gateway/providers/apply-provider-specific-logic.ts
index 1669bdcce7..f6f1751633 100644
--- a/apps/web/src/lib/ai-gateway/providers/apply-provider-specific-logic.ts
+++ b/apps/web/src/lib/ai-gateway/providers/apply-provider-specific-logic.ts
@@ -19,9 +19,12 @@ import {
   isOpus5Model,
 } from '@/lib/ai-gateway/providers/anthropic.constants';
 import { OpenRouterInferenceProviderIdSchema } from '@/lib/ai-gateway/providers/openrouter/inference-provider-id';
-import { applyMoonshotModelSettings, isKimiModel } from '@/lib/ai-gateway/providers/moonshotai';
+import {
+  applyMoonshotModelSettings,
+  isKimiModel,
+  KIMI_CURRENT_MODEL_ID,
+} from '@/lib/ai-gateway/providers/moonshotai';
 import { isGlmModel } from '@/lib/ai-gateway/providers/zai';
-import { PERPLEXITY_KIMI_PUBLIC_ID } from '@/lib/ai-gateway/providers/partner/constants';
 import { isMinimaxModel } from '@/lib/ai-gateway/providers/minimax';
 import {
   ReasoningDetailsTransform,
@@ -217,7 +220,7 @@ export function applyAnthropicThinkingDefault(
   const defaultsToThinking =
     (isMinimaxModel(requestedModel) && requestedModel.includes('m3')) ||
     requestedModel === 'z-ai/glm-5.2' ||
-    requestedModel === PERPLEXITY_KIMI_PUBLIC_ID;
+    requestedModel === KIMI_CURRENT_MODEL_ID;
   if (
     defaultsToThinking &&
     requestToMutate.kind === 'messages' &&
@@ -235,12 +238,11 @@ export function removeUnsupportedRequestServiceTier(
   kiloExclusiveModel: KiloExclusiveModel | null
 ) {
   const customPricing = getCustomPricing(requestedModel);
-  const reason =
-    customPricing && !customPricing.fallbackOnly
-      ? 'non-fallback custom pricing'
-      : kiloExclusiveModel && !kiloExclusiveModel.flags.includes('flex')
-        ? 'non-Flex Kilo-exclusive model'
-        : null;
+  const reason = customPricing
+    ? 'custom pricing'
+    : kiloExclusiveModel && !kiloExclusiveModel.flags.includes('flex')
+      ? 'non-Flex Kilo-exclusive model'
+      : null;
   const serviceTier = requestToMutate.body.service_tier;
   if (!reason || serviceTier === undefined) {
     return;
diff --git a/apps/web/src/lib/ai-gateway/providers/partner/constants.ts b/apps/web/src/lib/ai-gateway/providers/partner/constants.ts
deleted file mode 100644
index 0f35001577..0000000000
--- a/apps/web/src/lib/ai-gateway/providers/partner/constants.ts
+++ /dev/null
@@ -1,2 +0,0 @@
-// Partner routing stays pinned when the current public models change.
-export const PERPLEXITY_KIMI_PUBLIC_ID = 'moonshotai/kimi-k3';
diff --git a/apps/web/src/lib/ai-gateway/providers/partner/pricing.ts b/apps/web/src/lib/ai-gateway/providers/partner/pricing.ts
deleted file mode 100644
index a24b724319..0000000000
--- a/apps/web/src/lib/ai-gateway/providers/partner/pricing.ts
+++ /dev/null
@@ -1,24 +0,0 @@
-import type { PricingTiers } from '@/lib/ai-gateway/providers/kilo-exclusive-model';
-import { PERPLEXITY_KIMI_PUBLIC_ID } from '@/lib/ai-gateway/providers/partner/constants';
-
-type PartnerPricing = {
-  fallbackOnly: true;
-  pricing: PricingTiers;
-};
-
-export const partnerPricingByModelId = {
-  [PERPLEXITY_KIMI_PUBLIC_ID]: {
-    fallbackOnly: true,
-    pricing: [
-      {
-        start_context_length: 0,
-        pricing: {
-          prompt_per_million: 3,
-          completion_per_million: 15,
-          input_cache_read_per_million: 0.3,
-          input_cache_write_per_million: null,
-        },
-      },
-    ],
-  },
-} satisfies Record<string, PartnerPricing>;
diff --git a/apps/web/src/lib/ai-gateway/providers/partner/providers.test.ts b/apps/web/src/lib/ai-gateway/providers/partner/providers.test.ts
deleted file mode 100644
index 866fa1f006..0000000000
--- a/apps/web/src/lib/ai-gateway/providers/partner/providers.test.ts
+++ /dev/null
@@ -1,80 +0,0 @@
-import { describe, expect, test } from '@jest/globals';
-
-import type { GatewayRequest } from '@/lib/ai-gateway/providers/openrouter/types';
-import { PERPLEXITY_KIMI_PUBLIC_ID } from '@/lib/ai-gateway/providers/partner/constants';
-import { PERPLEXITY_KIMI_PROVIDER } from '@/lib/ai-gateway/providers/partner/providers';
-import {
-  ReasoningDetailsTransform,
-  type TransformRequestContext,
-} from '@/lib/ai-gateway/providers/types';
-
-describe('Perplexity Kimi provider', () => {
-  test('supports the chat completions API', () => {
-    expect(`${PERPLEXITY_KIMI_PROVIDER.apiUrl}/chat/completions`).toBe(
-      'https://api.perplexity.ai/router/v1/chat/completions'
-    );
-    expect(PERPLEXITY_KIMI_PROVIDER.supportedChatApis).toEqual(['chat_completions']);
-  });
-
-  test('enables the reasoning details response transform', () => {
-    expect(PERPLEXITY_KIMI_PROVIDER.responseTransforms).toBe(
-      ReasoningDetailsTransform.ReasoningContent
-    );
-  });
-
-  test('hardwires the upstream model and removes provider and user settings', async () => {
-    const request: GatewayRequest = {
-      kind: 'chat_completions',
-      body: {
-        model: PERPLEXITY_KIMI_PUBLIC_ID,
-        messages: [{ role: 'user', content: 'hello' }],
-        provider: { order: ['perplexity'] },
-        user: 'user-id',
-      },
-    };
-
-    await PERPLEXITY_KIMI_PROVIDER.transformRequest({ request } as TransformRequestContext);
-
-    expect(request.body.model).toBe('perplexity/kimi-k3');
-    expect(request.body.provider).toBeUndefined();
-    expect(request.body.user).toBeUndefined();
-  });
-
-  test.each([
-    [{ enabled: false as const, effort: 'none' as const }, 'none'],
-    [{ enabled: true as const, effort: 'high' as const }, 'high'],
-  ])('maps reasoning %p to Perplexity effort %s', async (reasoning, expected) => {
-    const request: GatewayRequest = {
-      kind: 'chat_completions',
-      body: {
-        model: PERPLEXITY_KIMI_PUBLIC_ID,
-        messages: [{ role: 'user', content: 'hello' }],
-        reasoning,
-        reasoning_effort: 'low',
-      },
-    };
-
-    await PERPLEXITY_KIMI_PROVIDER.transformRequest({ request } as TransformRequestContext);
-
-    expect(request.body.reasoning_effort).toBe(expected);
-    expect(request.body.reasoning).toBeUndefined();
-  });
-
-  test.each(['none', 'low', undefined] as const)(
-    'preserves legacy reasoning effort %p when reasoning settings are absent',
-    async reasoning_effort => {
-      const request: GatewayRequest = {
-        kind: 'chat_completions',
-        body: {
-          model: PERPLEXITY_KIMI_PUBLIC_ID,
-          messages: [{ role: 'user', content: 'hello' }],
-          reasoning_effort,
-        },
-      };
-
-      await PERPLEXITY_KIMI_PROVIDER.transformRequest({ request } as TransformRequestContext);
-
-      expect(request.body.reasoning_effort).toBe(reasoning_effort);
-    }
-  );
-});
diff --git a/apps/web/src/lib/ai-gateway/providers/partner/providers.ts b/apps/web/src/lib/ai-gateway/providers/partner/providers.ts
deleted file mode 100644
index d231bb1bc8..0000000000
--- a/apps/web/src/lib/ai-gateway/providers/partner/providers.ts
+++ /dev/null
@@ -1,34 +0,0 @@
-import { getEnvVariable } from '@/lib/dotenvx';
-import {
-  isReasoningExplicitlyDisabled,
-  removeChatCompletionsToolNames,
-} from '@/lib/ai-gateway/providers/openrouter/request-helpers';
-import { ReasoningDetailsTransform, type Provider } from '@/lib/ai-gateway/providers/types';
-
-export const PERPLEXITY_KIMI_PROVIDER = {
-  id: 'perplexity',
-  apiUrl: 'https://api.perplexity.ai/router/v1',
-  apiUrlOverrides: {},
-  apiKey: getEnvVariable('PERPLEXITY_API_KEY'),
-  apiKeyHeader: null,
-  supportedChatApis: [
-    'chat_completions',
-    // 'messages', // supported, not tested
-    // 'responses', // supported, not tested
-  ],
-  responseTransforms: ReasoningDetailsTransform.ReasoningContent,
-  async transformRequest(context) {
-    context.request.body.model = 'perplexity/kimi-k3';
-    if (context.request.kind === 'chat_completions') {
-      context.request.body.reasoning_effort = isReasoningExplicitlyDisabled(context.request)
-        ? 'none'
-        : (context.request.body.reasoning?.effort ??
-          context.request.body.reasoning_effort ??
-          undefined);
-      delete context.request.body.reasoning;
-      removeChatCompletionsToolNames(context.request.body);
-    }
-    delete context.request.body.provider;
-    delete context.request.body.user;
-  },
-} as const satisfies Provider;
diff --git a/apps/web/src/lib/ai-gateway/providers/partner/routing.test.ts b/apps/web/src/lib/ai-gateway/providers/partner/routing.test.ts
deleted file mode 100644
index e98aa5d5c5..0000000000
--- a/apps/web/src/lib/ai-gateway/providers/partner/routing.test.ts
+++ /dev/null
@@ -1,159 +0,0 @@
-import { afterAll, beforeEach, describe, expect, it } from '@jest/globals';
-
-import type { GatewayRequest } from '@/lib/ai-gateway/providers/openrouter/types';
-import { PERPLEXITY_KIMI_PUBLIC_ID } from '@/lib/ai-gateway/providers/partner/constants';
-import { PERPLEXITY_KIMI_PROVIDER } from '@/lib/ai-gateway/providers/partner/providers';
-import {
-  selectPercentageRoutedPartnerProvider,
-  type PercentageRoutedPartnerInput,
-} from '@/lib/ai-gateway/providers/partner/routing';
-import type { RuntimeGatewayRoutingConfig } from '@/lib/ai-gateway/providers/routing-config';
-
-const originalPerplexityApiKey = PERPLEXITY_KIMI_PROVIDER.apiKey;
-
-function setApiKey(value: string) {
-  Object.defineProperty(PERPLEXITY_KIMI_PROVIDER, 'apiKey', {
-    value,
-    configurable: true,
-    enumerable: true,
-    writable: true,
-  });
-}
-
-beforeEach(() => {
-  setApiKey('test-perplexity-key');
-});
-
-afterAll(() => {
-  setApiKey(originalPerplexityApiKey);
-});
-
-function request(
-  kind: GatewayRequest['kind'] = 'chat_completions',
-  provider?: GatewayRequest['body']['provider']
-): GatewayRequest {
-  if (kind === 'responses') {
-    return { kind, body: { model: 'model', input: 'hello', provider } };
-  }
-  return {
-    kind,
-    body: {
-      model: 'model',
-      max_tokens: 1_024,
-      messages: [{ role: 'user', content: 'hello' }],
-      provider,
-    },
-  };
-}
-
-const routingConfig: RuntimeGatewayRoutingConfig = {
-  vercelPaid: 50,
-  vercelFree: 50,
-  vercelOptOutModels: new Set(),
-  perplexity: 100,
-};
-
-const defaultInput = {
-  requestedModel: PERPLEXITY_KIMI_PUBLIC_ID,
-  request: request(),
-  randomSeed: 'user-id',
-  sourceProviderId: 'openrouter',
-  hasUserByok: false,
-} satisfies PercentageRoutedPartnerInput;
-
-function selectPartner(
-  overrides: Partial<PercentageRoutedPartnerInput> = {},
-  config = routingConfig
-) {
-  return selectPercentageRoutedPartnerProvider({ ...defaultInput, ...overrides }, config);
-}
-
-describe('selectPercentageRoutedPartnerProvider', () => {
-  it('routes the exact Kimi model for chat completions requests', () => {
-    expect(selectPartner()).toBe(PERPLEXITY_KIMI_PROVIDER);
-  });
-
-  it.each(['openrouter', 'vercel'] as const)(
-    'does not route GLM 5.2 from %s to a direct partner',
-    sourceProviderId => {
-      expect(selectPartner({ requestedModel: 'z-ai/glm-5.2', sourceProviderId })).toBeNull();
-    }
-  );
-
-  it.each(['messages', 'responses'] as const)('does not route untested %s requests', kind => {
-    expect(selectPartner({ request: request(kind) })).toBeNull();
-  });
-
-  it('allows an empty provider object', () => {
-    expect(selectPartner({ request: request('chat_completions', {}) })).toBe(
-      PERPLEXITY_KIMI_PROVIDER
-    );
-  });
-
-  it('honors provider filters', () => {
-    expect(selectPartner({ request: request('chat_completions', { only: ['perplexity'] }) })).toBe(
-      PERPLEXITY_KIMI_PROVIDER
-    );
-    expect(selectPartner({ request: request('chat_completions', { ignore: ['openai'] }) })).toBe(
-      PERPLEXITY_KIMI_PROVIDER
-    );
-    expect(
-      selectPartner({ request: request('chat_completions', { only: ['openai'] }) })
-    ).toBeNull();
-    expect(
-      selectPartner({ request: request('chat_completions', { ignore: ['perplexity'] }) })
-    ).toBeNull();
-  });
-
-  it.each([{ data_collection: 'deny' } as const, { zdr: true } as const])(
-    'routes the ZDR model with data policy options %p',
-    provider => {
-      expect(selectPartner({ request: request('chat_completions', provider) })).toBe(
-        PERPLEXITY_KIMI_PROVIDER
-      );
-    }
-  );
-
-  it.each(['openrouter', 'vercel'] as const)(
-    'routes from managed provider %s',
-    sourceProviderId => {
-      expect(selectPartner({ sourceProviderId })).toBe(PERPLEXITY_KIMI_PROVIDER);
-    }
-  );
-
-  it.each(['perplexity', 'custom', 'direct-byok'] as const)(
-    'does not override provider %s',
-    sourceProviderId => {
-      expect(selectPartner({ sourceProviderId })).toBeNull();
-    }
-  );
-
-  it.each(['openrouter', 'vercel'] as const)(
-    'does not override user BYOK from %s',
-    sourceProviderId => {
-      expect(selectPartner({ sourceProviderId, hasUserByok: true })).toBeNull();
-    }
-  );
-
-  it.each(['', '  '])('does not route with an empty partner API key %p', apiKey => {
-    setApiKey(apiKey);
-
-    expect(selectPartner()).toBeNull();
-  });
-
-  it.each([`${PERPLEXITY_KIMI_PUBLIC_ID}-fast`, 'perplexity/kimi-k3'])(
-    'does not route non-exact model %s',
-    model => {
-      expect(selectPartner({ requestedModel: model })).toBeNull();
-    }
-  );
-
-  it('does not route at zero percent', () => {
-    const disabledConfig: RuntimeGatewayRoutingConfig = {
-      ...routingConfig,
-      perplexity: 0,
-    };
-
-    expect(selectPartner({}, disabledConfig)).toBeNull();
-  });
-});
diff --git a/apps/web/src/lib/ai-gateway/providers/partner/routing.ts b/apps/web/src/lib/ai-gateway/providers/partner/routing.ts
deleted file mode 100644
index 669889a37c..0000000000
--- a/apps/web/src/lib/ai-gateway/providers/partner/routing.ts
+++ /dev/null
@@ -1,74 +0,0 @@
-import type { GatewayRequest } from '@/lib/ai-gateway/providers/openrouter/types';
-import { PERPLEXITY_KIMI_PUBLIC_ID } from '@/lib/ai-gateway/providers/partner/constants';
-import { PERPLEXITY_KIMI_PROVIDER } from '@/lib/ai-gateway/providers/partner/providers';
-import { getRuntimeGatewayRoutingConfig } from '@/lib/ai-gateway/providers/routing-config';
-import {
-  passesRoutingPercentage,
-  type RoutingCohort,
-} from '@/lib/ai-gateway/providers/routing-percentage';
-import type { Provider } from '@/lib/ai-gateway/providers/types';
-
-type PartnerRoute = {
-  provider: Provider;
-  cohort: Exclude<RoutingCohort, 'vercel'>;
-};
-
-export type PercentageRoutedPartnerInput = {
-  requestedModel: string;
-  request: GatewayRequest;
-  randomSeed: string;
-  sourceProviderId: Provider['id'];
-  hasUserByok: boolean;
-};
-
-const PARTNER_ROUTES: Readonly<Record<string, PartnerRoute>> = {
-  [PERPLEXITY_KIMI_PUBLIC_ID]: {
-    provider: PERPLEXITY_KIMI_PROVIDER,
-    cohort: 'perplexity',
-  },
-};
-
-function isPartnerProviderAllowed(request: GatewayRequest, providerId: Provider['id']) {
-  const provider = request.body.provider;
-  return (
-    (!provider?.only || provider.only.includes(providerId)) &&
-    (!provider?.ignore || !provider.ignore.includes(providerId))
-  );
-}
-
-function getEligiblePartnerRoute(input: PercentageRoutedPartnerInput): PartnerRoute | null {
-  const { requestedModel, request, sourceProviderId, hasUserByok } = input;
-  const route = PARTNER_ROUTES[requestedModel];
-  if (
-    !route ||
-    route.provider.apiKey.trim().length === 0 ||
-    hasUserByok ||
-    (sourceProviderId !== 'vercel' && sourceProviderId !== 'openrouter') ||
-    !isPartnerProviderAllowed(request, route.provider.id) ||
-    !route.provider.supportedChatApis.includes(request.kind)
-  ) {
-    return null;
-  }
-  return route;
-}
-
-export async function getPercentageRoutedPartnerProvider(
-  input: PercentageRoutedPartnerInput
-): Promise<Provider | null> {
-  if (!getEligiblePartnerRoute(input)) return null;
-  const routingConfig = await getRuntimeGatewayRoutingConfig();
-  return selectPercentageRoutedPartnerProvider(input, routingConfig);
-}
-
-export function selectPercentageRoutedPartnerProvider(
-  input: PercentageRoutedPartnerInput,
-  routingConfig: Awaited<ReturnType<typeof getRuntimeGatewayRoutingConfig>>
-): Provider | null {
-  const route = getEligiblePartnerRoute(input);
-  if (!route) return null;
-
-  const percentage = routingConfig[route.cohort];
-  return passesRoutingPercentage(route.cohort, input.randomSeed, percentage)
-    ? route.provider
-    : null;
-}
diff --git a/apps/web/src/lib/ai-gateway/providers/provider-definitions.test.ts b/apps/web/src/lib/ai-gateway/providers/provider-definitions.test.ts
index 2174aaa7f0..ea9564d136 100644
--- a/apps/web/src/lib/ai-gateway/providers/provider-definitions.test.ts
+++ b/apps/web/src/lib/ai-gateway/providers/provider-definitions.test.ts
@@ -66,6 +66,5 @@ describe('tryGetProviderById', () => {
 
   test('does not claim dynamically constructed providers', () => {
     expect(tryGetProviderById('direct-byok')).toBeUndefined();
-    expect(tryGetProviderById('perplexity')).toBeUndefined();
   });
 });
diff --git a/apps/web/src/lib/ai-gateway/providers/routing-config.ts b/apps/web/src/lib/ai-gateway/providers/routing-config.ts
index 1ba5839b35..c248141fe4 100644
--- a/apps/web/src/lib/ai-gateway/providers/routing-config.ts
+++ b/apps/web/src/lib/ai-gateway/providers/routing-config.ts
@@ -1,6 +1,5 @@
 import { createCachedFetch } from '@/lib/cached-fetch';
 import {
-  DEFAULT_PERPLEXITY_PERCENTAGE,
   DEFAULT_VERCEL_PERCENTAGE,
   DEFAULT_VERCEL_PERCENTAGE_FREE,
   GatewayRoutingConfigSchema,
@@ -13,14 +12,12 @@ export type RuntimeGatewayRoutingConfig = {
   vercelPaid: number;
   vercelFree: number;
   vercelOptOutModels: ReadonlySet<string>;
-  perplexity: number;
 };
 
 const DEFAULT_RUNTIME_GATEWAY_ROUTING_CONFIG: RuntimeGatewayRoutingConfig = {
   vercelPaid: DEFAULT_VERCEL_PERCENTAGE,
   vercelFree: DEFAULT_VERCEL_PERCENTAGE_FREE,
   vercelOptOutModels: new Set(),
-  perplexity: DEFAULT_PERPLEXITY_PERCENTAGE,
 };
 
 export const getRuntimeGatewayRoutingConfig = createCachedFetch<RuntimeGatewayRoutingConfig>(
@@ -37,7 +34,6 @@ export const getRuntimeGatewayRoutingConfig = createCachedFetch<RuntimeGatewayRo
       vercelPaid: config.vercel_routing_percentage ?? DEFAULT_VERCEL_PERCENTAGE,
       vercelFree: config.vercel_routing_percentage_free ?? DEFAULT_VERCEL_PERCENTAGE_FREE,
       vercelOptOutModels: new Set(config.vercel_routing_opt_out_models),
-      perplexity: config.perplexity_routing_percentage ?? DEFAULT_PERPLEXITY_PERCENTAGE,
     };
   },
   60_000,
diff --git a/apps/web/src/lib/ai-gateway/providers/routing-percentage.ts b/apps/web/src/lib/ai-gateway/providers/routing-percentage.ts
index edda6cf8f4..d63031d4f8 100644
--- a/apps/web/src/lib/ai-gateway/providers/routing-percentage.ts
+++ b/apps/web/src/lib/ai-gateway/providers/routing-percentage.ts
@@ -1,9 +1,7 @@
 import { getRandomNumber } from '@/lib/ai-gateway/getRandomNumber';
 
-export type RoutingCohort = 'vercel' | 'perplexity';
-
 export function passesRoutingPercentage(
-  cohort: RoutingCohort,
+  cohort: 'vercel',
   randomSeed: string,
   routingPercentage: number
 ) {
diff --git a/apps/web/src/lib/ai-gateway/providers/types.ts b/apps/web/src/lib/ai-gateway/providers/types.ts
index f9da8827d2..063aa75749 100644
--- a/apps/web/src/lib/ai-gateway/providers/types.ts
+++ b/apps/web/src/lib/ai-gateway/providers/types.ts
@@ -17,7 +17,6 @@ export type ProviderId =
   | 'longcat'
   | 'martian'
   | 'mistral'
-  | 'perplexity'
   | 'streamlake'
   | 'vercel'
   | 'custom'
diff --git a/apps/web/src/lib/ai-gateway/providers/vercel/index.test.ts b/apps/web/src/lib/ai-gateway/providers/vercel/index.test.ts
index 9be95ff3ca..1cb8e328bf 100644
--- a/apps/web/src/lib/ai-gateway/providers/vercel/index.test.ts
+++ b/apps/web/src/lib/ai-gateway/providers/vercel/index.test.ts
@@ -306,7 +306,6 @@ describe('shouldRouteToVercel', () => {
         vercelPaid: 100,
         vercelFree: 100,
         vercelOptOutModels: new Set(options?.optOut ? ['anthropic/claude-sonnet-4.5'] : []),
-        perplexity: 0,
       })),
     }));
     jest.doMock('@/lib/ai-gateway/is-free-model', () => ({
diff --git a/apps/web/src/lib/ai-gateway/rewriteModelResponse.test.ts b/apps/web/src/lib/ai-gateway/rewriteModelResponse.test.ts
index 831a568396..67701b8c1c 100644
--- a/apps/web/src/lib/ai-gateway/rewriteModelResponse.test.ts
+++ b/apps/web/src/lib/ai-gateway/rewriteModelResponse.test.ts
@@ -1370,25 +1370,6 @@ describe('rewriteModelResponse', () => {
     });
   });
 
-  test('preserves cost for models with fallback-only custom pricing', async () => {
-    const result = await rewriteModelResponse({
-      response: jsonResponse({
-        model: 'moonshotai/kimi-k3',
-        usage: { cost: 0.5, cost_details: { upstream_inference_cost: 0.4 }, is_byok: false },
-      }),
-      model: 'moonshotai/kimi-k3',
-      providerId: 'openrouter',
-      kind: 'chat_completions',
-      logging: makeLogging(),
-      responseTransforms: null,
-    });
-
-    expect(await result.json()).toEqual({
-      model: 'moonshotai/kimi-k3',
-      usage: { cost: 0.5, cost_details: { upstream_inference_cost: 0.4 }, is_byok: false },
-    });
-  });
-
   test('processes paid-model responses when request logging is enabled', async () => {
     mockedOptIn.mockResolvedValueOnce(true);
     const result = await rewriteModelResponse({
diff --git a/apps/web/src/lib/ai-gateway/rewriteModelResponse.ts b/apps/web/src/lib/ai-gateway/rewriteModelResponse.ts
index 645b91dd70..7a668a8556 100644
--- a/apps/web/src/lib/ai-gateway/rewriteModelResponse.ts
+++ b/apps/web/src/lib/ai-gateway/rewriteModelResponse.ts
@@ -934,8 +934,7 @@ export async function rewriteModelResponse({
   const customPricing = getCustomPricing(model);
   const requiresCostRemoval =
     (providerId === 'openrouter' || providerId === 'vercel') &&
-    (isKiloExclusiveFreeModel(model) ||
-      (customPricing !== undefined && !customPricing.fallbackOnly));
+    (isKiloExclusiveFreeModel(model) || customPricing !== undefined);
 
   console.debug('[rewriteModelResponse] rewriting response for %s', model);
   const { vercel_request_id: vercelRequestId } = logging;
diff --git a/apps/web/src/routers/admin/gateway-config-router.ts b/apps/web/src/routers/admin/gateway-config-router.ts
index 2b5b7e1b0a..ad2735f05b 100644
--- a/apps/web/src/routers/admin/gateway-config-router.ts
+++ b/apps/web/src/routers/admin/gateway-config-router.ts
@@ -38,7 +38,6 @@ export const adminGatewayConfigRouter = createTRPCRouter({
       vercel_routing_percentage: input.vercel_routing_percentage,
       vercel_routing_percentage_free: input.vercel_routing_percentage_free,
       vercel_routing_opt_out_models: optOutModels,
-      perplexity_routing_percentage: input.perplexity_routing_percentage,
       updated_at: new Date().toISOString(),
       updated_by: ctx.user.id,
       updated_by_email: ctx.user.google_user_email,