Skip to content
EN
English 简体中文 soon 日本語 soon

OpenRouter

One API for hundreds of models

Visit official site

What OpenRouter is

OpenRouter is a router rather than a model vendor. One OpenAI-compatible API reaches a very large number of models from many providers, and it can route, degrade and balance requests so an application can switch models without rewriting integration code.

What you can do with it

  • Call many providers through one key
  • Switch models with minimal code change
  • Configure fallbacks when a model is unavailable
  • Run comparisons across models
  • Balance load across providers

Who it is for

  • Developers wanting provider flexibility
  • Applications needing fallback capacity
  • Teams comparing models at scale

What to watch out for

  • An intermediary sees every prompt, so review its logging and retention commitments
  • Each upstream provider's terms still apply, including data handling and what you may build
  • Routing can silently change which model answered, so log the actual model used
  • Model names and availability change often, which affects reproducibility

Pros & cons

✓ What we like

  • Enormous model choice from one integration
  • Fallback and routing built in
  • Drop-in compatibility with common clients

! What to watch out for

  • Intermediary handles your prompts
  • Upstream terms still apply
  • Silent model substitution possible

FAQ

How many models are reachable?

A very large catalogue across many providers is advertised; verify the specific ones you need.

Does it work with existing clients?

The API is OpenAI-compatible, so most existing clients work unchanged.

What should I log myself?

The model that actually served each request, since routing can change it.

Last reviewed: 2026-09-19

More LLM API platform tools

View all →

How we review