Quick questions
Short prompts, no tools. Goes to a fast, cheap model.
Models
Every lab.One key.
At the provider's price plus 5%. Charged only for the tokens a model actually used.
Models
0 models.
One key.
OpenAI, Anthropic, Google, DeepSeek, xAI, Meta, Mistral and more, at the provider's price plus 5%. Set the model to auto and the router picks the cheapest one that can handle each request.
"summarize this thread"
fastClaude Haiku 5.5
Claude Haiku 5.5Auto routing
Set the model to auto.
The router does the rest.
Code and tools
Code, long input, tool calls, images. Goes to a stronger model.
Hard reasoning
Proofs, plans, deep analysis. Goes to the top tier.
Prices
A few of the 337.
About one short chat per request. Exact prices are per token and show in the app.