GatewayInternalServerError With Gemini & Grok Models

AI SDK
1

I’ve been running some models via AI-Gateway and am running into the following errors:

  • 400 for xai/grok-4-fast-reasoning
  • 401 for google/gemini-2.5-flash-lite

My other models (claude/claude-3-5-sonnet-20241022, meta/llama-3.1-8b, openai/gpt-4o) all work fine, and I’m passing the exact same stuff through all of them.
What could be the issue?

Gemini Complete Log:

AI Gateway error: Error [GatewayInternalServerError]: Request had invalid authentication credentials. Expected OAuth 2 access token, login cookie or other valid authentication credential. See https://developers.google.com/identity/sign-in/web/devconsole-project.
    at async E (.next/server/chunks/[root-of-the-server]__642d5f6a._.js:44:206) {
  statusCode: 401,
  type: 'internal_server_error',
  [cause]: Error [AI_APICallError]: [object Object]
      at ignore-listed frames {
    cause: undefined,
    url: 'https://ai-gateway.vercel.sh/v1/ai/language-model',
    requestBodyValues: {
      maxOutputTokens: undefined,
      temperature: 0.8,
      topP: undefined,
      topK: undefined,
      presencePenalty: undefined,
      frequencyPenalty: undefined,
      stopSequences: undefined,
      seed: undefined,
      tools: undefined,
      toolChoice: undefined,
      responseFormat: undefined,
      prompt: [Array],
      providerOptions: undefined,
      headers: [Object]
    },
    statusCode: 401,
    responseHeaders: {
      'cache-control': 'public, max-age=0, must-revalidate',
      'content-type': 'application/json',
      date: 'Sun, 14 Dec 2025 21:16:34 GMT',
      server: 'Vercel',
      'strict-transport-security': 'max-age=63072000; includeSubDomains; preload',
      'transfer-encoding': 'chunked',
      vary: 'rsc, next-router-state-tree, next-router-prefetch, next-router-segment-prefetch',
      'x-matched-path': '/v1/ai/language-model',
      'x-vercel-cache': 'MISS',
      'x-vercel-id': 'sfo1:iad1::iad1::8zlhk-1765746994202-fac38fd6c188'
    },
    responseBody: '{"error":{"message":"Request had invalid authentication credentials. Expected OAuth 2 access token, login cookie or other valid authentication credential. See https://developers.google.com/identity/sign-in/web/devconsole-project.","type":"AI_APICallError","param":{"error":"Request had invalid authentication credentials. Expected OAuth 2 access token, login cookie or other valid authentication credential. See https://developers.google.com/identity/sign-in/web/devconsole-project.","statusCode":401,"name":"AI_APICallError","message":"Request had invalid authentication credentials. Expected OAuth 2 access token, login cookie or other valid authentication credential. See https://developers.google.com/identity/sign-in/web/devconsole-project.","url":"https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-lite:generateContent","isRetryable":false,"type":"AI_APICallError"}}}',
    isRetryable: false,
    data: { error: [Object] }
  }
}

Grok Complete Log:

2025-12-14 21:16:49.958 [info] [AI Gateway] Calling model: xai/grok-4-fast-reasoning Environment: production
2025-12-14 21:16:50.036 [error] AI Gateway error: Error [GatewayInternalServerError]: Bad Request
    at async E (.next/server/chunks/[root-of-the-server]__642d5f6a._.js:44:206) {
  statusCode: 400,
  type: 'internal_server_error',
  [cause]: Error [AI_APICallError]: [object Object]
      at ignore-listed frames {
    cause: undefined,
    url: 'https://ai-gateway.vercel.sh/v1/ai/language-model',
    requestBodyValues: {
      maxOutputTokens: undefined,
      temperature: 0.8,
      topP: undefined,
      topK: undefined,
      presencePenalty: undefined,
      frequencyPenalty: undefined,
      stopSequences: undefined,
      seed: undefined,
      tools: undefined,
      toolChoice: undefined,
      responseFormat: undefined,
      prompt: [Array],
      providerOptions: undefined,
      headers: [Object]
    },
    statusCode: 400,
    responseHeaders: {
      'cache-control': 'public, max-age=0, must-revalidate',
      'content-type': 'application/json',
      date: 'Sun, 14 Dec 2025 21:16:50 GMT',
      server: 'Vercel',
      'strict-transport-security': 'max-age=63072000; includeSubDomains; preload',
      'transfer-encoding': 'chunked',
      vary: 'rsc, next-router-state-tree, next-router-prefetch, next-router-segment-prefetch',
      'x-matched-path': '/v1/ai/language-model',
      'x-vercel-cache': 'MISS',
      'x-vercel-id': 'sfo1:iad1::iad1::8vhdd-1765747009969-a0ff312fc70d'
    },
    responseBody: '{"error":{"message":"Bad Request","type":"AI_APICallError","param":{"error":"Bad Request","statusCode":400,"name":"AI_APICallError","message":"Bad Request","url":"https://api.x.ai/v1/chat/completions","isRetryable":false,"type":"AI_APICallError"}}}',
    isRetryable: false,
    data: { error: [Object] }
  }
}```
4

Hi there, the Google request is an auth failure – are you using BYOK? Are you able to submit requests of other models or are all Gemini models failing auth?

On the x.ai request, can you provide a sample code snippet that produces the issue, since it reads as if it is specific to the prompt/params of the request, so we can help investigate and diagnose?