One API key across hundreds of models, routed by price or by preference.
What it does
On this page
OpenRouter is a single gateway to models from many vendors. One key and one request shape reach text, image and audio models, and the platform decides which provider serves each call.
Who it is for
Developers who want to compare models, or fall back between vendors, without opening an account and learning a new SDK for every one.
How it works
Requests name a model slug, and routing can prefer the cheapest provider, the fastest, or a shortlist you define. If a provider fails, the request can be retried elsewhere, and failed attempts are not charged under its zero-completion policy. Logs show spend per key and per model, budgets cap it, prompt caching cuts repeated input cost, and provider keys you already own can be attached. Tool calling depends on the underlying model, so it works only where that model supports it.
Pricing
The free tier covers a small set of free models with a daily request ceiling. Paid use adds a platform fee on top of each provider's token price, with different rates for standard, business and enterprise accounts.
Notable limits
Free models are rate limited and unsuitable for production traffic. Platform fees are extra on top of model pricing, and enterprise features such as SSO require a sales conversation.
Source: https://openrouter.ai/pricing