Vision

With Vision you can send images to an AI model and have them described or analyzed.

Usage

1. Attach an image (PNG, JPG, GIF, WebP), just like with Files.
2. Ask a question about it, for example "Describe this image" or "What does the screenshot say?".
3. Send it; the image is sent along with your question to a vision-capable model.

Automatic Routing

• There is no separate vision model selection required; the right model is chosen automatically.
• If the currently selected chat model is already vision-capable, it is used directly.

Image Optimization

Images are optimized automatically before sending (size and compression) to balance quality and cost. The output format is preserved (PNG stays PNG, JPEG stays JPEG), and an image is never made larger than the original.

Cost & Limits

• Up to 5 images per message (a shared limit with other attachments), each up to 10 MB.
• Frontier Vision is billed via credits; fast models do not consume credits.
• If image analysis fails, you get a clear message; no credits are charged.

Vision requires an active subscription. See Credits & Subscriptions and Privacy.