Available Models

We carefully evaluate the Large Language Models available on the market, selecting those that work best together for a wide range of applications. Our team curates models based on their unique strengths and specializations, ensuring we can address diverse use cases effectively. This thoughtful selection process ensures our router provides access to a complementary mix of models tailored to different needs.

At the moment, the router supports the following models, grouped by the provider whose key executes them:

Model names and aliases

Use the names above in the model and models parameters. Provider-specific names and dated versions (for example gpt-5.2-2025-12-11 or claude-sonnet-4-5) are accepted as aliases and normalized to the names above. auto is not a model: it tells AI Router to decide for the request.

Model lifecycle

Models go through three lifecycle states:

• Active models are routable and selectable.
• Deprecated models remain routable and selectable so existing integrations keep working, but they are penalized in the ranking and marked in the dashboard. Move to a successor before the sunset date.
• Removed models are no longer routable or selectable. Requests that still name a removed model fail, so move to a successor model before the sunset date.

New models that we add for a provider are picked up automatically when Automatically add new models from enabled providers is switched on for your team; see Choosing Models.