If you build or use software that sends requests to AI models, OpenRouter gives you one API for accessing models from multiple providers. An API is a way for software to send requests to another service and receive responses. OpenRouter sends requests from your software to the provider and handles billing through one account.
OpenRouter is a platform for finding and using models, not an AI model itself. Its OpenAI-compatible API lets developers connect with tools built for OpenAI by changing the API endpoint and key. Developers can try different providers without building a separate connection for each one.
How OpenRouter routes model requests
OpenRouter sends each request to a selected model and provider. Users can sort providers by price, throughput, or latency and set performance thresholds. If a provider is unavailable or rate-limited, fallback settings route the request to another eligible provider or model. The provider-routing controls let developers manage those choices.
For example, a developer could use one model for most requests and configure a fallback for interruptions. OpenRouter also offers an Auto Router that selects a model based on the request. Automatic selection means developers do not choose a model for each request. Specifying a model gives them more control over which model handles it.
The catalog includes models from providers such as OpenAI, Anthropic, Google, Meta, and Mistral. It supports text and multimodal models, which can work with inputs such as images and PDFs. The catalog changes, so check the live model list instead of relying on a fixed model count.
What OpenRouter costs
OpenRouter uses pay-as-you-go billing with credits, so users pay through one account instead of managing a separate bill for each provider. Each model’s usage price follows its provider’s rates. OpenRouter says it adds no markup to those rates. Credit purchases still incur a fee: its FAQ lists a 5.5% fee with a $0.80 minimum, and a 5% fee for cryptocurrency payments.
Some models are free to use, with limits tied to credit purchases. The FAQ gives a default daily limit for free-model API requests and says users can raise it by purchasing credits. These limits apply to free models; check the current terms before planning around them.
OpenRouter also supports Bring Your Own Key (BYOK), which lets users connect provider credentials they already have. The fee and allowance depend on the plan, and usage beyond the allowance incurs a 5% fee. When using BYOK, check both the provider’s bill and OpenRouter’s usage figures to understand the full cost.
What to check before using it
OpenRouter simplifies access and billing, but requests still depend on the selected model and provider. Model and provider prices, availability, capabilities, and data handling can differ. Test the models and routing settings against your actual workload, especially before relying on an automatic fallback or router for production requests.
OpenRouter says it does not log prompts and completions by default. However, some providers have different or unconfirmed logging policies. OpenRouter says it will not route requests to those providers unless the user enables the relevant setting in privacy controls. Review both OpenRouter’s privacy and logging details and the provider’s policies before sending sensitive information.