🤖 feat: GPT-5.4 and GPT-5.4-pro Context + Pricing (#12099)

* ✨ feat: Add support for new GPT-5.4 and GPT-5.4-pro models

- Introduced new token values and cache settings for 'gpt-5.4' and 'gpt-5.4-pro' in the API model configurations.
- Updated maximum output limits for the new models in the tokens utility.
- Included 'gpt-5.4' and 'gpt-5.4-pro' in the shared OpenAI models list for consistent access across the application.

* 🔧 update: Enhance GPT-5.4 and GPT-5.4-pro model configurations

- Refined token pricing and cache settings for 'gpt-5.4' and 'gpt-5.4-pro' in the API model configurations.
- Added tests for cache multipliers and maximum token limits for the new models.
- Updated shared OpenAI models list to include 'gpt-5.4-thinking' and added a note for verifying pricing before release.

* 🔧 update: Add clarification to token pricing for 'gpt-5.4-pro'

- Added a comment to the 'gpt-5.4-pro' model configuration in tokens.ts to specify that it shares the same token window as 'gpt-5.4', enhancing clarity for future reference.

api/models/tx.js

@@ -4,31 +4,18 @@ const defaultRate = 6;
 /**
  * Token Pricing Configuration
  *
- * IMPORTANT: Key Ordering for Pattern Matching
- * ============================================
- * The `findMatchingPattern` function iterates through object keys in REVERSE order
- * (last-defined keys are checked first) and uses `modelName.includes(key)` for matching.
+ * Pattern Matching
+ * ================
+ * `findMatchingPattern` (from @librechat/api) uses `modelName.includes(key)` and selects
+ * the LONGEST matching key. If a key's length equals the model name's length (exact match),
+ * it returns immediately. Definition order does NOT affect correctness.
  *
- * This means:
- * 1. BASE PATTERNS must be defined FIRST (e.g., "kimi", "moonshot")
- * 2. SPECIFIC PATTERNS must be defined AFTER their base patterns (e.g., "kimi-k2", "kimi-k2.5")
- *
- * Example ordering for Kimi models:
- *   kimi: { prompt: 0.6, completion: 2.5 },       // Base pattern - checked last
- *   'kimi-k2': { prompt: 0.6, completion: 2.5 },  // More specific - checked before "kimi"
- *   'kimi-k2.5': { prompt: 0.6, completion: 3.0 }, // Most specific - checked first
- *
- * Why this matters:
- * - Model name "kimi-k2.5" contains both "kimi" and "kimi-k2" as substrings
- * - If "kimi" were checked first, it would incorrectly match and return wrong pricing
- * - By defining specific patterns AFTER base patterns, they're checked first in reverse iteration
+ * Key ordering matters only for:
+ *   1. Performance: list older/less common models first so newer/common models
+ *      are found earlier in the reverse scan.
+ *   2. Same-length tie-breaking: the last-defined key wins on equal-length matches.
  *
  * This applies to BOTH `tokenValues` and `cacheTokenValues` objects.
- *
- * When adding new model families:
- * 1. Define the base/generic pattern first
- * 2. Define increasingly specific patterns after
- * 3. Ensure no pattern is a substring of another that should match differently
  */
 
 /**
@@ -151,6 +138,9 @@ const tokenValues = Object.assign(
     'gpt-5.1': { prompt: 1.25, completion: 10 },
     'gpt-5.2': { prompt: 1.75, completion: 14 },
     'gpt-5.3': { prompt: 1.75, completion: 14 },
+    'gpt-5.4': { prompt: 2.5, completion: 15 },
+    // TODO: gpt-5.4-pro pricing not yet officially published — verify before release
+    'gpt-5.4-pro': { prompt: 5, completion: 30 },
     'gpt-5-nano': { prompt: 0.05, completion: 0.4 },
     'gpt-5-mini': { prompt: 0.25, completion: 2 },
     'gpt-5-pro': { prompt: 15, completion: 120 },
@@ -322,7 +312,7 @@ const cacheTokenValues = {
   //   gpt-4o (incl. mini), o1 (incl. mini/preview): 50% off
   //   gpt-4.1 (incl. mini/nano), o3 (incl. mini), o4-mini: 75% off
   //   gpt-5.x (excl. pro variants): 90% off
-  //   gpt-5-pro, gpt-5.2-pro: no caching
+  //   gpt-5-pro, gpt-5.2-pro, gpt-5.4-pro: no caching
   'gpt-4o': { write: 2.5, read: 1.25 },
   'gpt-4o-mini': { write: 0.15, read: 0.075 },
   'gpt-4.1': { write: 2, read: 0.5 },
@@ -332,6 +322,7 @@ const cacheTokenValues = {
   'gpt-5.1': { write: 1.25, read: 0.125 },
   'gpt-5.2': { write: 1.75, read: 0.175 },
   'gpt-5.3': { write: 1.75, read: 0.175 },
+  'gpt-5.4': { write: 2.5, read: 0.25 },
   'gpt-5-mini': { write: 0.25, read: 0.025 },
   'gpt-5-nano': { write: 0.05, read: 0.005 },
   o1: { write: 15, read: 7.5 },

api/models/tx.spec.js

@@ -59,6 +59,17 @@ describe('getValueKey', () => {
     expect(getValueKey('openai/gpt-5.3')).toBe('gpt-5.3');
   });
 
+  it('should return "gpt-5.4" for model name containing "gpt-5.4"', () => {
+    expect(getValueKey('gpt-5.4')).toBe('gpt-5.4');
+    expect(getValueKey('gpt-5.4-thinking')).toBe('gpt-5.4');
+    expect(getValueKey('openai/gpt-5.4')).toBe('gpt-5.4');
+  });
+
+  it('should return "gpt-5.4-pro" for model name containing "gpt-5.4-pro"', () => {
+    expect(getValueKey('gpt-5.4-pro')).toBe('gpt-5.4-pro');
+    expect(getValueKey('openai/gpt-5.4-pro')).toBe('gpt-5.4-pro');
+  });
+
   it('should return "gpt-3.5-turbo-1106" for model name containing "gpt-3.5-turbo-1106"', () => {
     expect(getValueKey('gpt-3.5-turbo-1106-some-other-info')).toBe('gpt-3.5-turbo-1106');
     expect(getValueKey('openai/gpt-3.5-turbo-1106')).toBe('gpt-3.5-turbo-1106');
@@ -400,6 +411,33 @@ describe('getMultiplier', () => {
     );
   });
 
+  it('should return the correct multiplier for gpt-5.4', () => {
+    expect(getMultiplier({ model: 'gpt-5.4', tokenType: 'prompt' })).toBe(
+      tokenValues['gpt-5.4'].prompt,
+    );
+    expect(getMultiplier({ model: 'gpt-5.4', tokenType: 'completion' })).toBe(
+      tokenValues['gpt-5.4'].completion,
+    );
+    expect(getMultiplier({ model: 'gpt-5.4-thinking', tokenType: 'prompt' })).toBe(
+      tokenValues['gpt-5.4'].prompt,
+    );
+    expect(getMultiplier({ model: 'openai/gpt-5.4', tokenType: 'completion' })).toBe(
+      tokenValues['gpt-5.4'].completion,
+    );
+  });
+
+  it('should return the correct multiplier for gpt-5.4-pro', () => {
+    expect(getMultiplier({ model: 'gpt-5.4-pro', tokenType: 'prompt' })).toBe(
+      tokenValues['gpt-5.4-pro'].prompt,
+    );
+    expect(getMultiplier({ model: 'gpt-5.4-pro', tokenType: 'completion' })).toBe(
+      tokenValues['gpt-5.4-pro'].completion,
+    );
+    expect(getMultiplier({ model: 'openai/gpt-5.4-pro', tokenType: 'prompt' })).toBe(
+      tokenValues['gpt-5.4-pro'].prompt,
+    );
+  });
+
   it('should return the correct multiplier for gpt-4o', () => {
     const valueKey = getValueKey('gpt-4o-2024-08-06');
     expect(getMultiplier({ valueKey, tokenType: 'prompt' })).toBe(tokenValues['gpt-4o'].prompt);
@@ -1377,6 +1415,7 @@ describe('getCacheMultiplier', () => {
       'gpt-5.1',
       'gpt-5.2',
       'gpt-5.3',
+      'gpt-5.4',
       'gpt-5-mini',
       'gpt-5-nano',
       'o1',
@@ -1413,10 +1452,20 @@ describe('getCacheMultiplier', () => {
     expect(getCacheMultiplier({ model: 'gpt-5-pro', cacheType: 'write' })).toBeNull();
     expect(getCacheMultiplier({ model: 'gpt-5.2-pro', cacheType: 'read' })).toBeNull();
     expect(getCacheMultiplier({ model: 'gpt-5.2-pro', cacheType: 'write' })).toBeNull();
+    expect(getCacheMultiplier({ model: 'gpt-5.4-pro', cacheType: 'read' })).toBeNull();
+    expect(getCacheMultiplier({ model: 'gpt-5.4-pro', cacheType: 'write' })).toBeNull();
   });
 
   it('should have consistent 10% cache read pricing for gpt-5.x models', () => {
-    const gpt5CacheModels = ['gpt-5', 'gpt-5.1', 'gpt-5.2', 'gpt-5.3', 'gpt-5-mini', 'gpt-5-nano'];
+    const gpt5CacheModels = [
+      'gpt-5',
+      'gpt-5.1',
+      'gpt-5.2',
+      'gpt-5.3',
+      'gpt-5.4',
+      'gpt-5-mini',
+      'gpt-5-nano',
+    ];
     for (const model of gpt5CacheModels) {
       expect(cacheTokenValues[model].read).toBeCloseTo(cacheTokenValues[model].write * 0.1, 10);
     }

api/utils/tokens.spec.js

@@ -214,6 +214,25 @@ describe('getModelMaxTokens', () => {
     );
   });
 
+  test('should return correct tokens for gpt-5.4 matches', () => {
+    expect(getModelMaxTokens('gpt-5.4')).toBe(maxTokensMap[EModelEndpoint.openAI]['gpt-5.4']);
+    expect(getModelMaxTokens('gpt-5.4-thinking')).toBe(
+      maxTokensMap[EModelEndpoint.openAI]['gpt-5.4'],
+    );
+    expect(getModelMaxTokens('openai/gpt-5.4')).toBe(
+      maxTokensMap[EModelEndpoint.openAI]['gpt-5.4'],
+    );
+  });
+
+  test('should return correct tokens for gpt-5.4-pro matches', () => {
+    expect(getModelMaxTokens('gpt-5.4-pro')).toBe(
+      maxTokensMap[EModelEndpoint.openAI]['gpt-5.4-pro'],
+    );
+    expect(getModelMaxTokens('openai/gpt-5.4-pro')).toBe(
+      maxTokensMap[EModelEndpoint.openAI]['gpt-5.4-pro'],
+    );
+  });
+
   test('should return correct tokens for Anthropic models', () => {
     const models = [
       'claude-2.1',
@@ -495,6 +514,8 @@ describe('getModelMaxTokens', () => {
       'gpt-5.1',
       'gpt-5.2',
       'gpt-5.3',
+      'gpt-5.4',
+      'gpt-5.4-pro',
       'gpt-5-mini',
       'gpt-5-nano',
       'gpt-5-pro',
@@ -804,6 +825,12 @@ describe('matchModelName', () => {
     expect(matchModelName('gpt-5.3-2025-03-01')).toBe('gpt-5.3');
   });
 
+  it('should return the closest matching key for gpt-5.4 matches', () => {
+    expect(matchModelName('openai/gpt-5.4')).toBe('gpt-5.4');
+    expect(matchModelName('gpt-5.4-thinking')).toBe('gpt-5.4');
+    expect(matchModelName('gpt-5.4-pro')).toBe('gpt-5.4-pro');
+  });
+
   it('should return the input model name if no match is found - Google models', () => {
     expect(matchModelName('unknown-google-model', EModelEndpoint.google)).toBe(
       'unknown-google-model',

packages/api/src/utils/tokens.ts

@@ -55,6 +55,8 @@ const openAIModels = {
   'gpt-5.1': 400000,
   'gpt-5.2': 400000,
   'gpt-5.3': 400000,
+  'gpt-5.4': 272000, // standard context; 1M experimental available via API opt-in (2x rate)
+  'gpt-5.4-pro': 272000, // same window as gpt-5.4
   'gpt-5-mini': 400000,
   'gpt-5-nano': 400000,
   'gpt-5-pro': 400000,
@@ -361,6 +363,8 @@ export const modelMaxOutputs = {
   'gpt-5.1': 128000,
   'gpt-5.2': 128000,
   'gpt-5.3': 128000,
+  'gpt-5.4': 128000,
+  'gpt-5.4-pro': 128000,
   'gpt-5-mini': 128000,
   'gpt-5-nano': 128000,
   'gpt-5-pro': 128000,

packages/data-provider/src/config.ts

@@ -1101,6 +1101,10 @@ export const alternateName = {
 };
 
 const sharedOpenAIModels = [
+  'gpt-5.4',
+  // TODO: gpt-5.4-thinking may have separate reasoning token pricing — verify before release
+  'gpt-5.4-thinking',
+  'gpt-5.4-pro',
   'gpt-5.1',
   'gpt-5.1-chat-latest',
   'gpt-5.1-codex',
@@ -1276,6 +1280,7 @@ export const visionModels = [
   'o4-mini',
   'o3',
   'o1',
+  'gpt-5',
   'gpt-4.1',
   'gpt-4.5',
   'llava',