Use model native image gen

Ideas Feature Requests
, ,
8

Hey @wangyy. You already started a separate thread for this, so let’s keep the quality discussion there so it doesn’t split across two places: GenerateImage dropped from Pro (~1.6MB photoreal) to Flash-class (~300KB CGI) overnight. I already replied there in detail and went through your measurements.

In short, what you’re seeing isn’t a silent Flash fallback and it isn’t a downgrade due to a rate limit. The model behind the GenerateImage tool is chosen on our side and can’t be selected in the app, and the default model can change over time. Your earlier images used the old default, and everything from around 08:00 Beijing time on September 9 used the new one. The change in file size and texture detail is from the default model switch, not from your prompts and not from the chat model.

A couple notes about the tool itself:

  • The GenerateImage schema only accepts description, filename, reference_image_paths, and aspect_ratio. Extra args like model, image_size, quality aren’t forwarded and are just ignored, you’re right about that.
  • The chat model Grok 4.6, Composer 2.5, or a sub-agent doesn’t affect which image model serves the request.

I’ve shared your before and after with the team that owns image generation as feedback on the new default.