Enable GPT-5.6 OpenRouter promotion

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

@@ -16,3 +16,20 @@ export const GPT_CURRENT_VERCEL_MODEL_ID = GPT_CURRENT_MODEL_ID;
 export const GPT_MINI_CURRENT_MODEL_ID = 'openai/gpt-5.4-mini';
 
 export const GPT_MINI_CURRENT_VERCEL_MODEL_ID = GPT_MINI_CURRENT_MODEL_ID;
+
+// OpenRouter BYOK must be disabled for these models so requests use the discounted endpoint.
+export const ENABLE_OPENROUTER_GPT56_PROMO = true;
+
+export const OPENROUTER_GPT56_PROMO_MODEL_IDS = [
+  'openai/gpt-5.6-terra',
+  'openai/gpt-5.6-terra-pro',
+  'openai/gpt-5.6-luna',
+  'openai/gpt-5.6-luna-pro',
+] as const;
+
+export function isOpenRouterGpt56PromoModel(modelId: string): boolean {
+  return (
+    ENABLE_OPENROUTER_GPT56_PROMO &&
+    OPENROUTER_GPT56_PROMO_MODEL_IDS.some(promoModelId => promoModelId === modelId)
+  );
+}

apps/web/src/lib/ai-gateway/providers/openrouter/index.test.ts

@@ -2,6 +2,7 @@ import { describe, it, expect, beforeEach, afterEach, jest } from '@jest/globals
 import {
   formatName,
   getEnhancedOpenRouterModels,
+  getModelDisplayPricing,
   getOpenRouterTranscriptionModels,
   shouldSuppressOpenRouterModel,
   undoPricingDiscount,
@@ -19,6 +20,7 @@ import {
 import type { KiloExclusiveModel } from '@/lib/ai-gateway/providers/kilo-exclusive-model';
 import { isFableModel } from '@/lib/ai-gateway/providers/anthropic.constants';
 import { KILO_AUTO_EFFICIENT_MODEL } from '@/lib/ai-gateway/auto-model';
+import { OPENROUTER_GPT56_PROMO_MODEL_IDS } from '@/lib/ai-gateway/providers/openai';
 
 jest.mock('@/lib/ai-gateway/providers/gateway-models-cache', () => ({
   getOpenRouterModelsMetadataFromDatabase: jest.fn(() => Promise.resolve({})),
@@ -150,6 +152,17 @@ describe('formatName', () => {
     });
     expect(formatName(model, NOT_PREFERRED)).toBe('OpenRouter Test Model ($$$$)');
   });
+
+  it.each(OPENROUTER_GPT56_PROMO_MODEL_IDS)('prefers the 50% off marker for %s', modelId => {
+    const recentlyCreated = Math.floor(Date.now() / 1000) - 24 * 3600;
+    const model = buildModel({
+      id: modelId,
+      created: recentlyCreated,
+      expiration_date: '2026-07-01',
+      pricing: { prompt: '0.00001', completion: '0' },
+    });
+    expect(formatName(model, 0)).toBe('Test Model (50% off)');
+  });
 });
 
 describe('undoPricingDiscount', () => {
@@ -191,6 +204,27 @@ describe('undoPricingDiscount', () => {
   });
 });
 
+describe('OpenRouter GPT-5.6 promotion', () => {
+  const discountedPricing = {
+    prompt: '0.000001',
+    completion: '0.000004',
+    input_cache_read: '0.0000001',
+    discount: 0.5,
+  };
+
+  it.each(OPENROUTER_GPT56_PROMO_MODEL_IDS)('preserves discounted pricing for %s', modelId => {
+    expect(getModelDisplayPricing(modelId, discountedPricing)).toBe(discountedPricing);
+  });
+
+  it('continues to undo endpoint discounts for other models', () => {
+    expect(getModelDisplayPricing('openai/gpt-5.6-sol', discountedPricing)).toEqual({
+      prompt: '0.000002000000',
+      completion: '0.000008000000',
+      input_cache_read: '0.000000200000',
+    });
+  });
+});
+
 describe('shouldSuppressOpenRouterModel', () => {
   it('does not suppress disabled paid Kilo-exclusive models returned by OpenRouter', () => {
     expect(disabledPaidModel.pricing).not.toBeNull();

apps/web/src/lib/ai-gateway/providers/openrouter/index.ts

@@ -28,6 +28,7 @@ import { getTerminalBenchSummaries, terminalBenchFor } from '@/lib/model-stats/t
 import { isFreeNemotronModel, NVIDIA_TRIAL_TOS } from '@/lib/ai-gateway/providers/nvidia';
 import { applyCustomPricingToModel } from '@/lib/ai-gateway/custom-pricing';
 import { addMonths } from 'date-fns';
+import { isOpenRouterGpt56PromoModel } from '@/lib/ai-gateway/providers/openai';
 
 // Re-export from shared module for backwards compatibility
 export { normalizeModelId } from '@/lib/ai-gateway/model-utils';
@@ -80,6 +81,7 @@ export function formatName(model: OpenRouterModel, preferredIndex: number) {
     model.id.startsWith('openrouter/') && !model.name.includes('OpenRouter')
       ? 'OpenRouter ' + model.name
       : model.name;
+  if (isOpenRouterGpt56PromoModel(model.id)) return name + ' (50% off)';
   const promptPrice = Number.parseFloat(model.pricing.prompt);
   const isExpensive = Number.isFinite(promptPrice) && promptPrice >= 0.00001; // Opus 4.8 Fast price
   if (isExpensive) return name + ' ($$$$)';
@@ -118,6 +120,14 @@ export function undoPricingDiscount(pricing: EndpointPricing): EndpointPricing {
   return result;
 }
 
+export function getModelDisplayPricing(
+  modelId: string,
+  pricing: EndpointPricing | undefined
+): EndpointPricing | undefined {
+  if (!pricing) return undefined;
+  return isOpenRouterGpt56PromoModel(modelId) ? pricing : undoPricingDiscount(pricing);
+}
+
 export function shouldSuppressOpenRouterModel(model: KiloExclusiveModel): boolean {
   return model.status !== 'disabled' || model.pricing === null;
 }
@@ -151,7 +161,7 @@ async function enhancedModelList(models: OpenRouterModel[]) {
                 normalizeInferenceProviderId(e.tag) ===
                 normalizeInferenceProviderId(preferredProvider)
             )?.pricing);
-        const pricing = rawPricing ? undoPricingDiscount(rawPricing) : rawPricing;
+        const pricing = getModelDisplayPricing(model.id, rawPricing);
         const terminalBench = terminalBenchFor(summaries, model.id);
         return {
           ...model,

apps/web/src/lib/ai-gateway/providers/vercel/index.test.ts

@@ -1,13 +1,15 @@
-import { describe, it, expect } from '@jest/globals';
+import { describe, it, expect, jest } from '@jest/globals';
 import {
   convertProviderOptions,
   getAnthropicProviderOptionsForVercel,
   getVercelInferenceProvidersExcludingIgnored,
   hasCompatibleVercelInferenceProvider,
   passesVercelRoutingPercentage,
+  shouldRouteToVercel,
 } from '@/lib/ai-gateway/providers/vercel';
 import { getRandomNumber } from '@/lib/ai-gateway/getRandomNumber';
 import type { GatewayRequest } from '@/lib/ai-gateway/providers/openrouter/types';
+import { OPENROUTER_GPT56_PROMO_MODEL_IDS } from '@/lib/ai-gateway/providers/openai';
 
 describe('getAnthropicProviderOptionsForVercel', () => {
   it('maps chat completion verbosity to Anthropic effort', () => {
@@ -154,3 +156,22 @@ describe('passesVercelRoutingPercentage', () => {
     expect(seedsInFinalBucket.some(seed => !passesVercelRoutingPercentage(seed, 99.9))).toBe(true);
   });
 });
+
+describe('OpenRouter GPT-5.6 promotion', () => {
+  it.each(OPENROUTER_GPT56_PROMO_MODEL_IDS)('does not route %s to Vercel', async model => {
+    const debugSpy = jest.spyOn(console, 'debug').mockImplementation(() => {});
+    const request: GatewayRequest = {
+      kind: 'chat_completions',
+      body: {
+        model,
+        messages: [{ role: 'user', content: 'hello' }],
+      },
+    };
+
+    await expect(shouldRouteToVercel(model, request, 'promo-test')).resolves.toBe(false);
+    expect(debugSpy).toHaveBeenCalledWith(
+      `[shouldRouteToVercel] routing ${model} to OpenRouter for the GPT-5.6 promotion`
+    );
+    debugSpy.mockRestore();
+  });
+});

apps/web/src/lib/ai-gateway/providers/vercel/index.ts

@@ -27,6 +27,7 @@ import {
   getVercelModelsFromRedis,
 } from '@/lib/ai-gateway/providers/gateway-models-cache';
 import type { AnthropicProviderOptions } from '@ai-sdk/anthropic';
+import { isOpenRouterGpt56PromoModel } from '@/lib/ai-gateway/providers/openai';
 
 type VercelRoutingPercentages = {
   paid: number;
@@ -94,6 +95,13 @@ export async function shouldRouteToVercel(
   request: GatewayRequest,
   randomSeed: string
 ) {
+  if (isOpenRouterGpt56PromoModel(requestedModel)) {
+    console.debug(
+      `[shouldRouteToVercel] routing ${requestedModel} to OpenRouter for the GPT-5.6 promotion`
+    );
+    return false;
+  }
+
   console.debug('[shouldRouteToVercel] randomizing user to either OpenRouter or Vercel');
   const percentages = await getVercelRoutingPercentages();
   const routingPercentage = (await isFreeModel(requestedModel))