🧠 feat: Add Claude Opus 4.8 Support (#13380)

* feat: add Claude Opus 4.8 support

* fix: omit sampling params for Claude Opus 4.8

* fix: flatten Bedrock beta header merge

* fix: strip Bedrock sampling params for Opus 4.8

.env.example

@@ -171,7 +171,7 @@ PROXY=
 #============#
 
 ANTHROPIC_API_KEY=user_provided
-# ANTHROPIC_MODELS=claude-opus-4-7,claude-sonnet-4-6,claude-opus-4-6,claude-opus-4-20250514,claude-3-7-sonnet-20250219,claude-3-5-sonnet-20241022,claude-3-5-haiku-20241022,claude-3-opus-20240229,claude-3-sonnet-20240229,claude-3-haiku-20240307
+# ANTHROPIC_MODELS=claude-opus-4-8,claude-opus-4-7,claude-sonnet-4-6,claude-opus-4-6,claude-opus-4-20250514,claude-3-7-sonnet-20250219,claude-3-5-sonnet-20241022,claude-3-5-haiku-20241022,claude-3-opus-20240229,claude-3-sonnet-20240229,claude-3-haiku-20240307
 # ANTHROPIC_REVERSE_PROXY=
 
 # Set to true to use Anthropic models through Google Vertex AI instead of direct API
@@ -237,8 +237,8 @@ ANTHROPIC_API_KEY=user_provided
 # BEDROCK_AWS_BEARER_TOKEN=yourBedrockApiKey
 
 # Note: This example list is not meant to be exhaustive. If omitted, all known, supported model IDs will be included for you.
-# BEDROCK_AWS_MODELS=anthropic.claude-opus-4-7,anthropic.claude-sonnet-4-6,anthropic.claude-opus-4-6-v1,anthropic.claude-3-5-sonnet-20240620-v1:0,meta.llama3-1-8b-instruct-v1:0
-# Cross-region inference model IDs: us.anthropic.claude-opus-4-7,us.anthropic.claude-sonnet-4-6,us.anthropic.claude-opus-4-6-v1,global.anthropic.claude-opus-4-6-v1
+# BEDROCK_AWS_MODELS=anthropic.claude-opus-4-8,anthropic.claude-opus-4-7,anthropic.claude-sonnet-4-6,anthropic.claude-opus-4-6-v1,anthropic.claude-3-5-sonnet-20240620-v1:0,meta.llama3-1-8b-instruct-v1:0
+# Cross-region inference model IDs: us.anthropic.claude-opus-4-8,us.anthropic.claude-opus-4-7,us.anthropic.claude-sonnet-4-6,us.anthropic.claude-opus-4-6-v1,global.anthropic.claude-opus-4-6-v1
 
 # See all Bedrock model IDs here: https://docs.aws.amazon.com/bedrock/latest/userguide/model-ids.html#model-ids-arns
 

api/utils/tokens.spec.js

@@ -1459,6 +1459,37 @@ describe('Claude Model Tests', () => {
     });
   });
 
+  it('should return correct context length for Claude Opus 4.8 (1M)', () => {
+    expect(getModelMaxTokens('claude-opus-4-8', EModelEndpoint.anthropic)).toBe(
+      maxTokensMap[EModelEndpoint.anthropic]['claude-opus-4-8'],
+    );
+    expect(getModelMaxTokens('claude-opus-4-8')).toBe(
+      maxTokensMap[EModelEndpoint.anthropic]['claude-opus-4-8'],
+    );
+  });
+
+  it('should return correct max output tokens for Claude Opus 4.8 (128K)', () => {
+    const { getModelMaxOutputTokens } = require('@librechat/api');
+    expect(getModelMaxOutputTokens('claude-opus-4-8', EModelEndpoint.anthropic)).toBe(
+      maxOutputTokensMap[EModelEndpoint.anthropic]['claude-opus-4-8'],
+    );
+  });
+
+  it('should match model names correctly for Claude Opus 4.8', () => {
+    const modelVariations = [
+      'claude-opus-4-8',
+      'claude-opus-4-8-20260528',
+      'claude-opus-4-8-latest',
+      'anthropic/claude-opus-4-8',
+      'claude-opus-4-8/anthropic',
+      'claude-opus-4-8-preview',
+    ];
+
+    modelVariations.forEach((model) => {
+      expect(matchModelName(model, EModelEndpoint.anthropic)).toBe('claude-opus-4-8');
+    });
+  });
+
   it('should return correct context length for Claude Sonnet 4.6 (1M)', () => {
     expect(getModelMaxTokens('claude-sonnet-4-6', EModelEndpoint.anthropic)).toBe(
       maxTokensMap[EModelEndpoint.anthropic]['claude-sonnet-4-6'],

client/src/locales/en/translation.json

@@ -223,7 +223,7 @@
   "com_endpoint_agent": "Agent",
   "com_endpoint_agent_placeholder": "Please select an Agent",
   "com_endpoint_ai": "AI",
-  "com_endpoint_anthropic_effort": "Controls how much computational effort Claude applies. Lower effort saves tokens and reduces latency; higher effort produces more thorough responses. 'Max' enables the deepest reasoning (Opus 4.6 only).",
+  "com_endpoint_anthropic_effort": "Controls how much computational effort Claude applies. Lower effort saves tokens and reduces latency; higher effort produces more thorough responses. 'Max' enables the deepest reasoning on supported adaptive-thinking models.",
   "com_endpoint_anthropic_maxoutputtokens": "Maximum number of tokens that can be generated in the response. Specify a lower value for shorter responses and a higher value for longer responses. Note: models may stop before reaching this maximum.",
   "com_endpoint_anthropic_prompt_cache": "Prompt caching allows reusing large context or instructions across API calls, reducing costs and latency",
   "com_endpoint_anthropic_temp": "Ranges from 0 to 1. Use temp closer to 0 for analytical / multiple choice, and closer to 1 for creative and generative tasks. We recommend altering this or Top P but not both.",

librechat.example.yaml

@@ -401,6 +401,7 @@ endpoints:
   #     # Option 1: Simple array (legacy format - model name = deployment name)
   #     # Use this if you want the technical model IDs to show in the UI
   #     # models:
+  #     #   - "claude-opus-4-8"
   #     #   - "claude-sonnet-4-6"
   #     #   - "claude-3-7-sonnet-20250219"
   #     #   - "claude-3-5-sonnet-v2@20241022"
@@ -411,6 +412,8 @@ endpoints:
   #     # The deploymentName is the actual Vertex AI model ID used for API calls
   #     # You can use friendly names (avoid spaces for cleaner YAML) or technical IDs as keys
   #     models:
+  #       claude-opus-4.8:
+  #         deploymentName: claude-opus-4-8
   #       claude-opus-4.5:
   #         deploymentName: claude-opus-4-5@20251101
   #       claude-sonnet-4:

packages/api/src/endpoints/anthropic/llm.spec.ts

@@ -888,7 +888,7 @@ describe('getLLMConfig', () => {
         expect(opus64kResult.llmConfig.maxTokens).toBe(64000);
 
         // opus-4-6+ get 128K
-        const opus128kModels = ['claude-opus-4-7', 'claude-opus-4-10'];
+        const opus128kModels = ['claude-opus-4-7', 'claude-opus-4-8', 'claude-opus-4-10'];
         opus128kModels.forEach((model) => {
           const result = getLLMConfig('test-key', {
             modelOptions: { model },
@@ -1066,6 +1066,22 @@ describe('getLLMConfig', () => {
         expect(thinking.display).toBe('summarized');
       });
 
+      it('should omit sampling parameters for Opus 4.8', () => {
+        const result = getLLMConfig('test-key', {
+          modelOptions: {
+            model: 'claude-opus-4-8',
+            thinking: false,
+            temperature: 0.7,
+            topP: 0.9,
+            topK: 40,
+          },
+        });
+
+        expect(result.llmConfig).not.toHaveProperty('temperature');
+        expect(result.llmConfig).not.toHaveProperty('topP');
+        expect(result.llmConfig).not.toHaveProperty('topK');
+      });
+
       it('should NOT set thinking.display for pre-Opus-4.7 adaptive models', () => {
         const pre47Models = ['claude-opus-4-6', 'claude-sonnet-4-6'];
 

packages/api/src/endpoints/anthropic/llm.ts

@@ -3,6 +3,7 @@ import { logger } from '@librechat/data-schemas';
 import { AnthropicClientOptions } from '@librechat/agents';
 import {
   anthropicSettings,
+  omitsSamplingParameters,
   removeNullishValues,
   ThinkingDisplay,
   AuthKeys,
@@ -177,9 +178,12 @@ function getLLMConfig(
     );
   }
 
+  const resolvedModel = requestOptions.model ?? mergedOptions.model;
+  const shouldOmitSamplingParameters = omitsSamplingParameters(resolvedModel);
+
   requestOptions = configureReasoning(requestOptions, systemOptions);
 
-  if (supportsAdaptiveThinking(mergedOptions.model)) {
+  if (supportsAdaptiveThinking(resolvedModel)) {
     if (
       systemOptions.effort &&
       (systemOptions.effort as string) !== '' &&
@@ -204,8 +208,8 @@ function getLLMConfig(
 
   const hasActiveThinking = requestOptions.thinking != null;
   const isThinkingModel =
-    /claude-3[-.]7/.test(mergedOptions.model) || supportsAdaptiveThinking(mergedOptions.model);
-  if (!isThinkingModel || !hasActiveThinking) {
+    /claude-3[-.]7/.test(resolvedModel) || supportsAdaptiveThinking(resolvedModel);
+  if (!shouldOmitSamplingParameters && (!isThinkingModel || !hasActiveThinking)) {
     requestOptions.topP = mergedOptions.topP;
     requestOptions.topK = mergedOptions.topK;
   }
@@ -293,6 +297,12 @@ function getLLMConfig(
     });
   }
 
+  if (shouldOmitSamplingParameters) {
+    delete requestOptions.temperature;
+    delete requestOptions.topP;
+    delete requestOptions.topK;
+  }
+
   const tools = [];
 
   if (enableWebSearch) {

packages/api/src/utils/tokens.ts

@@ -146,6 +146,7 @@ const anthropicModels = {
   'claude-sonnet-4-6': 1000000,
   'claude-opus-4-6': 1000000,
   'claude-opus-4-7': 1000000,
+  'claude-opus-4-8': 1000000,
 };
 
 const deepseekModels = {
@@ -394,6 +395,7 @@ const anthropicMaxOutputs = {
   'claude-opus-4-5': 64000,
   'claude-opus-4-6': 128000,
   'claude-opus-4-7': 128000,
+  'claude-opus-4-8': 128000,
   'claude-3.5-sonnet': 8192,
   'claude-3-5-sonnet': 8192,
   'claude-3.7-sonnet': 128000,

packages/data-provider/specs/bedrock.spec.ts

@@ -2,6 +2,7 @@ import { ThinkingDisplay } from '../src/schemas';
 import {
   BEDROCK_OUTPUT_128K_BETA,
   supportsAdaptiveThinking,
+  omitsSamplingParameters,
   omitsThinkingByDefault,
   resolveThinkingDisplay,
   bedrockOutputParser,
@@ -222,7 +223,7 @@ describe('omitsThinkingByDefault', () => {
     expect(omitsThinkingByDefault('us.anthropic.claude-opus-4-7')).toBe(true);
   });
 
-  test('returns true for claude-opus-4-8 (future Opus 4.x)', () => {
+  test('returns true for claude-opus-4-8', () => {
     expect(omitsThinkingByDefault('claude-opus-4-8')).toBe(true);
   });
 
@@ -238,6 +239,11 @@ describe('omitsThinkingByDefault', () => {
     expect(omitsThinkingByDefault('claude-opus-4-6')).toBe(false);
   });
 
+  test('returns false for base Opus 4 snapshot IDs', () => {
+    expect(omitsThinkingByDefault('claude-opus-4-20250514')).toBe(false);
+    expect(omitsThinkingByDefault('anthropic.claude-opus-4-20250514-v1:0')).toBe(false);
+  });
+
   test('returns false for claude-opus-4-5', () => {
     expect(omitsThinkingByDefault('claude-opus-4-5')).toBe(false);
   });
@@ -264,6 +270,36 @@ describe('omitsThinkingByDefault', () => {
   });
 });
 
+describe('omitsSamplingParameters', () => {
+  test('returns true for Opus 4.7+ models', () => {
+    const models = [
+      'claude-opus-4-7',
+      'claude-opus-4-8',
+      'anthropic.claude-opus-4-8',
+      'us.anthropic.claude-opus-4-8',
+      'claude-opus-5',
+    ];
+
+    models.forEach((model) => {
+      expect(omitsSamplingParameters(model)).toBe(true);
+    });
+  });
+
+  test('returns false for older Opus and non-Opus models', () => {
+    const models = [
+      'claude-opus-4-20250514',
+      'anthropic.claude-opus-4-20250514-v1:0',
+      'claude-opus-4-1-20250805',
+      'claude-opus-4-6',
+      'claude-sonnet-4-7',
+    ];
+
+    models.forEach((model) => {
+      expect(omitsSamplingParameters(model)).toBe(false);
+    });
+  });
+});
+
 describe('resolveThinkingDisplay', () => {
   test('returns "summarized" for Opus 4.7 when explicit is auto/null/undefined', () => {
     expect(resolveThinkingDisplay('claude-opus-4-7', ThinkingDisplay.auto)).toBe('summarized');
@@ -541,6 +577,47 @@ describe('bedrockInputParser', () => {
       expect(additionalFields.effort).toBeUndefined();
     });
 
+    test('should strip sampling parameters for Opus 4.8 Bedrock models', () => {
+      const input = {
+        model: 'anthropic.claude-opus-4-8',
+        temperature: 0.7,
+        topP: 0.9,
+        topK: 40,
+        top_p: 0.8,
+        additionalModelRequestFields: {
+          custom_flag: true,
+          temperature: 0.5,
+          topP: 0.95,
+          top_k: 20,
+        },
+      };
+      const result = bedrockInputParser.parse(input) as Record<string, unknown>;
+      const additionalFields = result.additionalModelRequestFields as Record<string, unknown>;
+      expect(result.temperature).toBeUndefined();
+      expect(result.topP).toBeUndefined();
+      expect(additionalFields.temperature).toBeUndefined();
+      expect(additionalFields.topP).toBeUndefined();
+      expect(additionalFields.top_p).toBeUndefined();
+      expect(additionalFields.top_k).toBeUndefined();
+      expect(additionalFields.custom_flag).toBe(true);
+      expect(additionalFields.thinking).toEqual({ type: 'adaptive', display: 'summarized' });
+      expect(additionalFields.anthropic_beta).toEqual(BEDROCK_CLAUDE_4_BETAS);
+    });
+
+    test('should preserve sampling parameters for Opus 4.6 Bedrock models', () => {
+      const input = {
+        model: 'anthropic.claude-opus-4-6-v1',
+        temperature: 0.7,
+        topP: 0.9,
+        topK: 40,
+      };
+      const result = bedrockInputParser.parse(input) as Record<string, unknown>;
+      const additionalFields = result.additionalModelRequestFields as Record<string, unknown>;
+      expect(result.temperature).toBe(0.7);
+      expect(result.topP).toBe(0.9);
+      expect(additionalFields.top_k).toBe(40);
+    });
+
     test('should set thinking.display to "summarized" so Opus 4.7 returns reasoning blocks', () => {
       const input = {
         model: 'anthropic.claude-opus-4-7',

packages/data-provider/src/bedrock.ts

@@ -68,14 +68,14 @@ type AnthropicInput = BedrockConverseInput & {
 
 /** Extracts opus major/minor version from both naming formats */
 function parseOpusVersion(model: string): { major: number; minor: number } | null {
-  const nameFirst = model.match(/claude-opus[-.]?(\d+)(?:[-.](\d+))?/);
+  const nameFirst = model.match(/claude-opus[-.]?(\d+)(?:[-.](\d{1,2})(?!\d))?/);
   if (nameFirst) {
     return {
       major: parseInt(nameFirst[1], 10),
       minor: nameFirst[2] != null ? parseInt(nameFirst[2], 10) : 0,
     };
   }
-  const numFirst = model.match(/claude-(\d+)(?:[-.](\d+))?-opus/);
+  const numFirst = model.match(/claude-(\d+)(?:[-.](\d{1,2})(?!\d))?-opus/);
   if (numFirst) {
     return {
       major: parseInt(numFirst[1], 10),
@@ -136,6 +136,14 @@ export function omitsThinkingByDefault(model: string): boolean {
   return false;
 }
 
+export function omitsSamplingParameters(model: string): boolean {
+  const opus = parseOpusVersion(model);
+  if (opus && (opus.major > 4 || (opus.major === 4 && opus.minor >= 7))) {
+    return true;
+  }
+  return false;
+}
+
 /** Checks if a model has a 1M context window (Sonnet 4.6+, Opus 4.6+, Opus 5+) */
 export function supportsContext1m(model: string): boolean {
   const sonnet = parseSonnetVersion(model);
@@ -177,11 +185,12 @@ function getBedrockAnthropicBetaHeaders(model: string): string[] {
 }
 
 function mergeBedrockAnthropicBetaHeaders(existing: unknown, generated: string[]): string[] {
-  const existingValues: unknown[] = Array.isArray(existing)
-    ? existing
-    : typeof existing === 'string'
-      ? [existing]
-      : [];
+  let existingValues: unknown[] = [];
+  if (Array.isArray(existing)) {
+    existingValues = existing;
+  } else if (typeof existing === 'string') {
+    existingValues = [existing];
+  }
 
   const betaHeaders = new Set<string>();
 
@@ -308,6 +317,8 @@ export const bedrockInputParser = s.tConversationSchema
 
     const additionalFields: Record<string, unknown> = {};
     const typedData = data as Record<string, unknown>;
+    const shouldOmitSamplingParameters =
+      typeof typedData.model === 'string' && omitsSamplingParameters(typedData.model);
 
     Object.entries(typedData).forEach(([key, value]) => {
       if (!knownKeys.includes(key)) {
@@ -436,6 +447,24 @@ export const bedrockInputParser = s.tConversationSchema
         delete amrf.reasoning_config;
         delete amrf.reasoning_effort;
       }
+
+      if (shouldOmitSamplingParameters) {
+        delete amrf.temperature;
+        delete amrf.topP;
+        delete amrf.top_p;
+        delete amrf.topK;
+        delete amrf.top_k;
+      }
+    }
+
+    if (shouldOmitSamplingParameters) {
+      delete typedData.temperature;
+      delete typedData.topP;
+      delete additionalFields.temperature;
+      delete additionalFields.topP;
+      delete additionalFields.top_p;
+      delete additionalFields.topK;
+      delete additionalFields.top_k;
     }
 
     /** Default promptCache for claude and nova models, if not defined */

packages/data-provider/src/config.ts

@@ -1542,6 +1542,7 @@ const sharedOpenAIModels = [
 ];
 
 const sharedAnthropicModels = [
+  'claude-opus-4-8',
   'claude-opus-4-7',
   'claude-sonnet-4-6',
   'claude-opus-4-6',
@@ -1565,6 +1566,7 @@ const sharedAnthropicModels = [
 ];
 
 export const bedrockModels = [
+  'anthropic.claude-opus-4-8',
   'anthropic.claude-opus-4-7',
   'anthropic.claude-sonnet-4-6',
   'anthropic.claude-opus-4-6-v1',

packages/data-schemas/src/methods/tx.spec.ts

@@ -2337,6 +2337,46 @@ describe('Claude Model Tests', () => {
       cacheTokenValues['claude-opus-4-7'].read,
     );
   });
+
+  it('should return correct prompt and completion rates for Claude Opus 4.8', () => {
+    expect(getMultiplier({ model: 'claude-opus-4-8', tokenType: 'prompt' })).toBe(
+      tokenValues['claude-opus-4-8'].prompt,
+    );
+    expect(getMultiplier({ model: 'claude-opus-4-8', tokenType: 'completion' })).toBe(
+      tokenValues['claude-opus-4-8'].completion,
+    );
+  });
+
+  it('should handle Claude Opus 4.8 model name variations', () => {
+    const modelVariations = [
+      'claude-opus-4-8',
+      'claude-opus-4-8-20260528',
+      'claude-opus-4-8-latest',
+      'anthropic/claude-opus-4-8',
+      'claude-opus-4-8/anthropic',
+      'claude-opus-4-8-preview',
+    ];
+
+    modelVariations.forEach((model) => {
+      const valueKey = getValueKey(model);
+      expect(valueKey).toBe('claude-opus-4-8');
+      expect(getMultiplier({ model, tokenType: 'prompt' })).toBe(
+        tokenValues['claude-opus-4-8'].prompt,
+      );
+      expect(getMultiplier({ model, tokenType: 'completion' })).toBe(
+        tokenValues['claude-opus-4-8'].completion,
+      );
+    });
+  });
+
+  it('should return correct cache rates for Claude Opus 4.8', () => {
+    expect(getCacheMultiplier({ model: 'claude-opus-4-8', cacheType: 'write' })).toBe(
+      cacheTokenValues['claude-opus-4-8'].write,
+    );
+    expect(getCacheMultiplier({ model: 'claude-opus-4-8', cacheType: 'read' })).toBe(
+      cacheTokenValues['claude-opus-4-8'].read,
+    );
+  });
 });
 
 describe('Premium Token Pricing', () => {
@@ -2357,7 +2397,12 @@ describe('Premium Token Pricing', () => {
   });
 
   it('should not apply premium pricing to Claude 1M GA models', () => {
-    const claudeModels = ['claude-opus-4-6', 'claude-opus-4-7', 'claude-sonnet-4-6'];
+    const claudeModels = [
+      'claude-opus-4-6',
+      'claude-opus-4-7',
+      'claude-opus-4-8',
+      'claude-sonnet-4-6',
+    ];
     claudeModels.forEach((model) => {
       expect(premiumTokenValues[model]).toBeUndefined();
       expect(getPremiumRate(model, 'prompt', wellAboveThreshold)).toBeNull();

packages/data-schemas/src/methods/tx.ts

@@ -158,6 +158,7 @@ export const tokenValues: Record<string, { prompt: number; completion: number }>
     'claude-opus-4-5': { prompt: 5, completion: 25 },
     'claude-opus-4-6': { prompt: 5, completion: 25 },
     'claude-opus-4-7': { prompt: 5, completion: 25 },
+    'claude-opus-4-8': { prompt: 5, completion: 25 },
     'claude-sonnet-4': { prompt: 3, completion: 15 },
     'claude-sonnet-4-5': { prompt: 3, completion: 15 },
     'claude-sonnet-4-6': { prompt: 3, completion: 15 },
@@ -294,6 +295,7 @@ export const cacheTokenValues: Record<string, { write: number; read: number }> =
   'claude-opus-4-5': { write: 6.25, read: 0.5 },
   'claude-opus-4-6': { write: 6.25, read: 0.5 },
   'claude-opus-4-7': { write: 6.25, read: 0.5 },
+  'claude-opus-4-8': { write: 6.25, read: 0.5 },
   '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 },