xAI's Grok Imagine Image 2.0 Preview is now live on Vercel AI Gateway. The model handles dense, multi-part visuals, specifically infographics, posters, and title screens, by co-planning typography and layout so small text stays legible and structure holds.
It supports image editing via prompt.images, where you pass an image alongside an instruction and the model changes only what you specified. Output resolution is configurable at 1k or 2k under providerOptions.xai, and the n parameter allows multiple images per call. Integration is one line: set model to xai/grok-imagine-image-2.0-preview and call generateImage from the AI SDK.
The implementation details, particularly how resolution tiers interact with generation consistency across repeated edits, are worth reading in full. A live demo runs at imagine.vercel.sh. The full list of supported image models on AI Gateway is linked from the changelog.
[READ ORIGINAL →]