@felixarntz
Yes, it occurs when I send another message after Gemini generates an image. Here's how I am able to reproduce it:

Message 1 (works)

// Call context:
"providerOptions": {
    "google": {
        "imageConfig": {
        "imageSize": "512",
        "aspectRatio": "1:1"
        }
    },
},

"messages": [{
    "role": "user",
    "content": "Draw an airplane"
}]
      
streamText({
    model: this.registry.languageModel('google:gemini-3.1-flash-image-preview'),
    providerOptions,
    messages,
    temperature: 1,
    maxRetries: 3,
  })

Message 2 (fails)

// Call context
"providerOptions": {
    "google": {
        "imageConfig": {
        "imageSize": "512",
        "aspectRatio": "1:1"
        }
    },
},

"messages": [{
    "role": "user",
    "content": "Draw an airplane"
   },
   {
    "role": "assistant",
    "content": [
    {
        // The previously generated image
        "type": "file",
        "data": "<Buffer length=453670>",
        "filename": "generated-cd2bb276-1772526367292-1.png",
        "mediaType": "image/png"
    }
    ]
},
{
    "role": "user",
    "content": "Place a giraffe on the airplane"
}]

streamText({
    model: this.registry.languageModel('google:gemini-3.1-flash-image-preview'),
    providerOptions,
    messages,
    temperature: 1,
    maxRetries: 3,
  })

It gives me the following error:

"responseBody": "{\n  \"error\": {\n    \"code\": 400,\n    \"message\": \"Image part is missing a thought_signature in content position 2, part position 1. Please refer to https://ai.google.dev/gemini-api/docs/thought-signatures#model-behavior for more details.\",\n    \"status\": \"INVALID_ARGUMENT\"\n  }\n}\n",
"isRetryable": false,
"data": {
"error": {
    "code": 400,
    "message": "Image part is missing a thought_signature in content position 2, part position 1. Please refer to https://ai.google.dev/gemini-api/docs/thought-signatures#model-behavior for more details.",
    "status": "INVALID_ARGUMENT"
}
},
"code": 400

This flow already works fine with gemini-2.5-flash-image so it seems to be related to the new model.

AI SDK versions I use:

"@ai-sdk/google": "^3.0.36",
"ai": "^6.0.108",

Google: Previous images may lack thought signature #12988

Description

Metadata

Metadata

Assignees

Labels

Type

Fields

Projects

Milestone

Relationships

Development

Issue actions