feat(models): add pricing property to KiloExclusiveModel

kilo-code-bot[bot] · kilo-code-bot[bot] · commit bfc7968c721a · 2026-04-08T20:26:27.000Z
- Add Pricing and Usage types to KiloExclusiveModel
- Remove 'free' flag from KiloExclusiveModelFlag
- pricing: null indicates a free model (replaces 'free' flag)
- Update qwen3.6 model with tiered pricing (low/high tier based on 256k input threshold)
- Update all existing models to use pricing: null
- Update isKiloExclusiveFreeModel to check pricing === null
- Update convertFromKiloExclusiveModel to expose pricing in metadata
- Add tests for qwen pricing calculation
diff --git a/apps/web/src/lib/models.test.ts b/apps/web/src/lib/models.test.ts
@@ -21,10 +21,10 @@ describe('isFreeModel', () => {
       expect(isFreeModel('openrouter/sonoma-sky-beta')).toBe(true);
     });
 
-    test('should return true for enabled Kilo exclusive models with free flag', () => {
-      // Test with known Kilo exclusive models that are enabled and have free flag
+    test('should return true for enabled Kilo exclusive models with no pricing', () => {
+      // Test with known Kilo exclusive models that are enabled and have no pricing (free)
       const enabledFreeModels = kiloExclusiveModels.filter(
-        m => m.status === 'public' && m.flags.includes('free')
+        m => m.status === 'public' && !m.pricing
       );
 
       // Should have at least some enabled free models
@@ -36,20 +36,23 @@ describe('isFreeModel', () => {
       }
     });
 
-    test('should return true for all Kilo exclusive models', () => {
-      // All kilo exclusive models currently have 'free' in their flags
-      // This test ensures isFreeModel returns true for all of them
-      for (const model of kiloExclusiveModels) {
-        if (model.status !== 'disabled') {
-          expect(isFreeModel(model.public_id)).toBe(true);
-        }
+    test('should return false for enabled Kilo exclusive models with pricing', () => {
+      // Models with pricing should NOT be free
+      const pricedModels = kiloExclusiveModels.filter(m => m.status !== 'disabled' && !!m.pricing);
+
+      for (const model of pricedModels) {
+        expect(isFreeModel(model.public_id)).toBe(false);
       }
     });
 
-    test('all Kilo exclusive models should have free flag', () => {
-      // Verify that all kilo exclusive models have the 'free' flag
+    test('all Kilo exclusive models should have either no pricing or valid pricing', () => {
+      // Verify that all kilo exclusive models have valid pricing structure
       for (const model of kiloExclusiveModels) {
-        expect(model.flags.includes('free')).toBe(true);
+        if (model.pricing) {
+          expect(typeof model.pricing.prompt_per_million).toBe('number');
+          expect(typeof model.pricing.completion_per_million).toBe('number');
+          expect(typeof model.pricing.calculate_mUsd).toBe('function');
+        }
       }
     });
 
diff --git a/apps/web/src/lib/models.ts b/apps/web/src/lib/models.ts
@@ -21,7 +21,7 @@ import { MINIMAX_CURRENT_MODEL_ID, minimax_m25_free_model } from '@/lib/provider
 import { KIMI_CURRENT_MODEL_ID } from '@/lib/providers/moonshotai';
 import { morph_warp_grep_free_model } from '@/lib/providers/morph';
 import { gpt_oss_20b_free_model } from '@/lib/providers/openai';
-import { qwen36_plus_free_model } from '@/lib/providers/qwen';
+import { qwen36_plus_model } from '@/lib/providers/qwen';
 import { grok_code_fast_1_optimized_free_model } from '@/lib/providers/xai';
 import { mimo_v2_omni_free_model, mimo_v2_pro_free_model } from '@/lib/providers/xiaomi';
 
@@ -69,7 +69,7 @@ export function isFreeModel(model: string): boolean {
 
 export function isKiloExclusiveFreeModel(model: string): boolean {
   return kiloExclusiveModels.some(
-    m => m.public_id === model && m.status !== 'disabled' && m.flags.includes('free')
+    m => m.public_id === model && m.status !== 'disabled' && !m.pricing
   );
 }
 
@@ -85,7 +85,7 @@ export const kiloExclusiveModels = [
   morph_warp_grep_free_model,
   grok_code_fast_1_optimized_free_model,
   seed_20_pro_free_model,
-  qwen36_plus_free_model,
+  qwen36_plus_model,
   trinity_large_thinking_free_model,
 ] as KiloExclusiveModel[];
 
@@ -99,6 +99,6 @@ function isOpenRouterStealthModel(model: string): boolean {
 
 export function isDeadFreeModel(model: string): boolean {
   return !!kiloExclusiveModels.find(
-    m => m.public_id === model && m.status === 'disabled' && m.flags.includes('free')
+    m => m.public_id === model && m.status === 'disabled' && !m.pricing
   );
 }
diff --git a/apps/web/src/lib/providers/arcee.ts b/apps/web/src/lib/providers/arcee.ts
@@ -8,8 +8,9 @@ export const trinity_large_thinking_free_model: KiloExclusiveModel = {
   context_length: 262_144,
   max_completion_tokens: 262_144,
   status: 'public',
-  flags: ['free', 'reasoning'],
+  flags: ['reasoning'],
   gateway: 'openrouter',
   internal_id: 'arcee-ai/trinity-large-thinking',
   inference_provider: 'arcee-ai',
+  pricing: null,
 };
diff --git a/apps/web/src/lib/providers/bytedance.ts b/apps/web/src/lib/providers/bytedance.ts
@@ -8,8 +8,9 @@ export const seed_20_pro_free_model: KiloExclusiveModel = {
   context_length: 256_000,
   max_completion_tokens: 128_000,
   status: 'public',
-  flags: ['free', 'reasoning', 'prompt_cache', 'vision'],
+  flags: ['reasoning', 'prompt_cache', 'vision'],
   gateway: 'bytedance',
   internal_id: 'seed-2-0-pro-260328',
   inference_provider: 'seed',
+  pricing: null,
 };
diff --git a/apps/web/src/lib/providers/corethink.ts b/apps/web/src/lib/providers/corethink.ts
@@ -8,8 +8,9 @@ export const corethink_free_model: KiloExclusiveModel = {
   context_length: 78_000,
   max_completion_tokens: 8192,
   status: 'public',
-  flags: ['free'],
+  flags: [],
   gateway: 'corethink',
   internal_id: 'corethink',
   inference_provider: 'corethink',
+  pricing: null,
 };
diff --git a/apps/web/src/lib/providers/kilo-exclusive-model.ts b/apps/web/src/lib/providers/kilo-exclusive-model.ts
@@ -1,7 +1,22 @@
 import type { OpenRouterInferenceProviderId } from '@/lib/providers/openrouter/inference-provider-id';
 import type { ProviderId } from '@/lib/providers/types';
 
-export type KiloExclusiveModelFlag = 'free' | 'reasoning' | 'prompt_cache' | 'vision';
+export type KiloExclusiveModelFlag = 'reasoning' | 'prompt_cache' | 'vision';
+
+export type Usage = {
+  inputTokens: number;
+  outputTokens: number;
+  cacheWriteTokens: number;
+  cacheHitTokens: number;
+};
+
+export type Pricing = {
+  prompt_per_million: number;
+  completion_per_million: number;
+  input_cache_read_per_million: number | null;
+  input_cache_write_per_million: number | null;
+  calculate_mUsd(usage: Usage, pricing: Pricing): number;
+};
 
 export type KiloExclusiveModel = {
   public_id: string;
@@ -14,9 +29,17 @@ export type KiloExclusiveModel = {
   gateway: ProviderId;
   internal_id: string;
   inference_provider: OpenRouterInferenceProviderId | null;
+  pricing: Pricing | null;
 };
 
+function formatPrice(price: number | null): string | undefined {
+  if (price === null) return undefined;
+  return price.toFixed(7);
+}
+
 export function convertFromKiloExclusiveModel(model: KiloExclusiveModel) {
+  const pricing = model.pricing;
+  const isFree = !pricing;
   return {
     id: model.public_id,
     canonical_slug: model.public_id,
@@ -33,13 +56,21 @@ export function convertFromKiloExclusiveModel(model: KiloExclusiveModel) {
       instruct_type: null,
     },
     pricing: {
-      prompt: '0.0000000',
-      completion: '0.0000000',
+      prompt: isFree
+        ? '0.0000000'
+        : (formatPrice(pricing.prompt_per_million / 1000000) ?? '0.0000000'),
+      completion: isFree
+        ? '0.0000000'
+        : (formatPrice(pricing.completion_per_million / 1000000) ?? '0.0000000'),
       request: '0',
       image: '0',
       web_search: '0',
       internal_reasoning: '0',
-      input_cache_read: model.flags.includes('prompt_cache') ? '0.00000000' : undefined,
+      input_cache_read: model.flags.includes('prompt_cache')
+        ? isFree
+          ? '0.00000000'
+          : (formatPrice(pricing.input_cache_read_per_million ?? 0 / 1000000) ?? '0.00000000')
+        : undefined,
     },
     top_provider: {
       context_length: model.context_length,
diff --git a/apps/web/src/lib/providers/minimax.ts b/apps/web/src/lib/providers/minimax.ts
@@ -8,10 +8,11 @@ export const minimax_m25_free_model: KiloExclusiveModel = {
   context_length: 204800,
   max_completion_tokens: 131072,
   status: 'hidden', // usable through kilo-auto/free
-  flags: ['free', 'reasoning', 'prompt_cache'],
+  flags: ['reasoning', 'prompt_cache'],
   gateway: 'openrouter',
   internal_id: 'minimax/minimax-m2.5',
   inference_provider: null,
+  pricing: null,
 };
 
 export function isMinimaxModel(model: string) {
diff --git a/apps/web/src/lib/providers/morph.ts b/apps/web/src/lib/providers/morph.ts
@@ -8,8 +8,9 @@ export const morph_warp_grep_free_model: KiloExclusiveModel = {
   context_length: 256000,
   max_completion_tokens: 32000,
   status: 'hidden',
-  flags: ['free'],
+  flags: [],
   gateway: 'morph',
   internal_id: 'morph-warp-grep-v2',
   inference_provider: 'morph',
+  pricing: null,
 };
diff --git a/apps/web/src/lib/providers/openai.ts b/apps/web/src/lib/providers/openai.ts
@@ -20,8 +20,9 @@ export const gpt_oss_20b_free_model: KiloExclusiveModel = {
   context_length: 131072,
   max_completion_tokens: 32768,
   status: 'hidden', // usable through kilo-auto
-  flags: ['free', 'reasoning'],
+  flags: ['reasoning'],
   gateway: 'openrouter',
   internal_id: 'openai/gpt-oss-20b',
   inference_provider: null,
+  pricing: null,
 };
diff --git a/apps/web/src/lib/providers/openrouter/sync-providers.ts b/apps/web/src/lib/providers/openrouter/sync-providers.ts
@@ -265,7 +265,7 @@ async function syncProviders() {
             provider_model_id: model.id,
             quantization: null,
             variant: 'default',
-            is_free: true,
+            is_free: !kfm.pricing,
             can_abort: true,
             max_prompt_tokens: model.top_provider.context_length,
             max_completion_tokens: model.top_provider.max_completion_tokens,
diff --git a/apps/web/src/lib/providers/qwen.test.ts b/apps/web/src/lib/providers/qwen.test.ts
@@ -0,0 +1,89 @@
+import { describe, test, expect } from '@jest/globals';
+import { calculateQwen36PlusCost } from './qwen';
+import type { Usage } from './kilo-exclusive-model';
+
+describe('calculateQwen36PlusCost', () => {
+  describe('tier 1 pricing (<= 256k total input) - 35% discounted rates', () => {
+    test('should calculate cost for basic usage within tier 1', () => {
+      const usage: Usage = {
+        inputTokens: 1000,
+        outputTokens: 500,
+        cacheWriteTokens: 0,
+        cacheHitTokens: 0,
+      };
+      // (1000 * 0.325 + 500 * 1.95) / 1_000_000 * 1_000_000 = 325 + 975 = 1300 microdollars
+      expect(calculateQwen36PlusCost(usage)).toBe(1300);
+    });
+
+    test('should calculate cost at exactly 256k threshold', () => {
+      const usage: Usage = {
+        inputTokens: 256 * 1024,
+        outputTokens: 0,
+        cacheWriteTokens: 0,
+        cacheHitTokens: 0,
+      };
+      // 262144 * 0.325 / 1_000_000 * 1_000_000 = 85196.8 -> rounded to 85197 microdollars
+      expect(calculateQwen36PlusCost(usage)).toBe(85197);
+    });
+
+    test('should include cache read and write costs', () => {
+      const usage: Usage = {
+        inputTokens: 1000,
+        outputTokens: 500,
+        cacheWriteTokens: 100,
+        cacheHitTokens: 200,
+      };
+      // (1000 * 0.325 + 500 * 1.95 + 100 * 0.40625 + 200 * 0.0325) / 1_000_000 * 1_000_000
+      // = 325 + 975 + 40.625 + 6.5 = 1347.125 -> rounded to 1347 microdollars
+      expect(calculateQwen36PlusCost(usage)).toBe(1347);
+    });
+  });
+
+  describe('tier 2 pricing (> 256k total input) - 35% discounted rates', () => {
+    test('should calculate cost for usage above tier 1 threshold', () => {
+      const usage: Usage = {
+        inputTokens: 300_000,
+        outputTokens: 1000,
+        cacheWriteTokens: 0,
+        cacheHitTokens: 0,
+      };
+      // (300000 * 1.3 + 1000 * 3.9) / 1_000_000 * 1_000_000 = 390000 + 3900 = 393900 microdollars
+      expect(calculateQwen36PlusCost(usage)).toBe(393900);
+    });
+
+    test('should calculate cost with cache at tier 2 rates', () => {
+      const usage: Usage = {
+        inputTokens: 300_000,
+        outputTokens: 1000,
+        cacheWriteTokens: 500,
+        cacheHitTokens: 1000,
+      };
+      // (300000 * 1.3 + 1000 * 3.9 + 500 * 1.625 + 1000 * 0.13) / 1_000_000 * 1_000_000
+      // = 390000 + 3900 + 812.5 + 130 = 394842.5 -> rounded to 394843 microdollars
+      expect(calculateQwen36PlusCost(usage)).toBe(394843);
+    });
+  });
+
+  describe('edge cases', () => {
+    test('should return 0 for zero usage', () => {
+      const usage: Usage = {
+        inputTokens: 0,
+        outputTokens: 0,
+        cacheWriteTokens: 0,
+        cacheHitTokens: 0,
+      };
+      expect(calculateQwen36PlusCost(usage)).toBe(0);
+    });
+
+    test('should round to nearest microdollar', () => {
+      const usage: Usage = {
+        inputTokens: 1,
+        outputTokens: 0,
+        cacheWriteTokens: 0,
+        cacheHitTokens: 0,
+      };
+      // 1 * 0.325 / 1_000_000 * 1_000_000 = 0.325 -> rounded to 0 microdollars
+      expect(calculateQwen36PlusCost(usage)).toBe(0);
+    });
+  });
+});
diff --git a/apps/web/src/lib/providers/qwen.ts b/apps/web/src/lib/providers/qwen.ts
@@ -1,15 +1,72 @@
-import type { KiloExclusiveModel } from '@/lib/providers/kilo-exclusive-model';
+import type { KiloExclusiveModel, Pricing, Usage } from '@/lib/providers/kilo-exclusive-model';
 
-export const qwen36_plus_free_model: KiloExclusiveModel = {
-  public_id: 'qwen/qwen3.6-plus:free',
-  display_name: 'Qwen: Qwen3.6 Plus (free)',
+const QWEN_256K_THRESHOLD = 256 * 1024;
+
+// Tier 1 pricing (<= 256k total input) - 35% lower than standard rates
+const qwenTier1Pricing: Pricing = {
+  prompt_per_million: 0.325,
+  completion_per_million: 1.95,
+  input_cache_read_per_million: 0.0325,
+  input_cache_write_per_million: 0.40625,
+  calculate_mUsd(usage: Usage, pricing: Pricing): number {
+    const inputCost = (usage.inputTokens * pricing.prompt_per_million) / 1000000;
+    const outputCost = (usage.outputTokens * pricing.completion_per_million) / 1000000;
+    const cacheReadCost =
+      pricing.input_cache_read_per_million !== null
+        ? (usage.cacheHitTokens * pricing.input_cache_read_per_million) / 1000000
+        : 0;
+    const cacheWriteCost =
+      pricing.input_cache_write_per_million !== null
+        ? (usage.cacheWriteTokens * pricing.input_cache_write_per_million) / 1000000
+        : 0;
+    return Math.round((inputCost + outputCost + cacheReadCost + cacheWriteCost) * 1000000);
+  },
+};
+
+// Tier 2 pricing (> 256k total input) - 35% lower than standard rates
+const qwenTier2Pricing: Pricing = {
+  prompt_per_million: 1.3,
+  completion_per_million: 3.9,
+  input_cache_read_per_million: 0.13,
+  input_cache_write_per_million: 1.625,
+  calculate_mUsd(usage: Usage, pricing: Pricing): number {
+    const inputCost = (usage.inputTokens * pricing.prompt_per_million) / 1000000;
+    const outputCost = (usage.outputTokens * pricing.completion_per_million) / 1000000;
+    const cacheReadCost =
+      pricing.input_cache_read_per_million !== null
+        ? (usage.cacheHitTokens * pricing.input_cache_read_per_million) / 1000000
+        : 0;
+    const cacheWriteCost =
+      pricing.input_cache_write_per_million !== null
+        ? (usage.cacheWriteTokens * pricing.input_cache_write_per_million) / 1000000
+        : 0;
+    return Math.round((inputCost + outputCost + cacheReadCost + cacheWriteCost) * 1000000);
+  },
+};
+
+export function calculateQwen36PlusCost(usage: Usage): number {
+  const totalInput = usage.inputTokens + usage.cacheWriteTokens + usage.cacheHitTokens;
+  const pricing = totalInput <= QWEN_256K_THRESHOLD ? qwenTier1Pricing : qwenTier2Pricing;
+  return pricing.calculate_mUsd(usage, pricing);
+}
+
+export const qwen36_plus_model: KiloExclusiveModel = {
+  public_id: 'qwen/qwen3.6-plus',
+  display_name: 'Qwen: Qwen3.6 Plus',
   description:
     'The Qwen3.6 native vision-language Plus series models demonstrate exceptional performance on par with the current state-of-the-art models, with a significant improvement in overall results compared to the 3.5 series. The models have been markedly enhanced in code-related capabilities such as agentic coding, front-end programming, and Vibe coding, as well as in multi-modal general object recognition, OCR, and object localization.',
   context_length: 1000000,
   max_completion_tokens: 65536,
   status: 'disabled',
-  flags: ['free', 'reasoning', 'prompt_cache', 'vision'],
+  flags: ['reasoning', 'vision'],
   gateway: 'alibaba',
   internal_id: 'qwen3.6-plus',
   inference_provider: 'alibaba',
+  pricing: {
+    prompt_per_million: 0.325,
+    completion_per_million: 1.95,
+    input_cache_read_per_million: 0.0325,
+    input_cache_write_per_million: 0.40625,
+    calculate_mUsd: calculateQwen36PlusCost,
+  },
 };
diff --git a/apps/web/src/lib/providers/xai.ts b/apps/web/src/lib/providers/xai.ts
diff --git a/apps/web/src/lib/providers/xiaomi.ts b/apps/web/src/lib/providers/xiaomi.ts
