The model index

Every (model, provider) pair — text LLMs plus image video audio generators — with price, context and capabilities. Blended $/1M is the effective cost of a typical 3:1 input:output request (what we rank text models by). Image/video/audio models are priced per unit by the provider (per image, per second) — not per token — so their token fields read “—”. Prices are provider-declared (via models.dev); measured evals are a later step. Machine: GET /v1/models (same filters incl. kind) · MCP: find_model.