Made for the way you work
Models and providers
Use the model that fits the work. Keep several providers ready in the same editor and change your choice as the task changes.
In practice
A useful place to start
Use a fast model for a small edit, then choose a stronger reasoning model for a difficult design review. Both work with the same project and tools.
A broad set of routes
Connect supported OpenAI, Anthropic, xAI, Google Gemini, Meta, and Fireworks models. Add reviewed Amazon Bedrock models or inference profiles, and signed-in Codex or Claude Code subscription models.
One model, several ways in
When several configured providers reach the same model, the picker groups them together. Your provider order chooses the ready route. You can see which route is in use instead of sorting through duplicate names.
Know what a model supports
Compare context size, image input, tools, reasoning, and prices in the model views. Edit your own skill ratings and descriptions, hide models you do not use, and set thinking levels where supported.
Get started
Open Settings, then Models. Add the relevant key or subscription, enable the models you want, and choose one in a chat. The site's Models page lists the current public catalog.
Good to know
Capabilities differ by model and route. Bedrock access depends on your account and region. Subscription routes currently handle text rather than image input; unknown capabilities are not a promise of support.