Every model.
One endpoint.
One API for 470 models. OpenRouter sends each request to a provider that can serve it, and retries on another if one fails.
OpenRouter in numbers
164TTokens last week
34×Growth in a year
470Models
112Providers
Tokens a week
Every week of the last year. The latest is lit.
Every model, one endpoint
Release date by blended price (three parts input to one part output), sized by context. The line is each quarter’s median.
Pick a lab to see only its models.
Weekly tokens from the public feed behind openrouter.ai/rankings; models and providers from openrouter.ai/api/v1.
What one key gets you
Every model, from every provider that serves it, behind one request. Pick a model to see its providers, prices and policies.
Agents run on OpenRouter
From the blog
Announcements, tutorials and case studies from the team.