OpenRouter

To begin using OpenRouter, create an account and add credits on the Credits page. Credits are deposits that you use for LLM inference, and they're deducted from your account as you make API requests. Once you have credits, you can either use the chat interface or create API keys to start using the API.

OpenRouter eliminates the complexity of managing multiple AI provider APIs. Instead of juggling separate API keys and authentication methods for OpenAI, Anthropic, Google, Meta, and others, you use a single API key to access their entire model catalog. The platform automatically handles authentication, billing, and error handling while providing automatic fallbacks to backup providers if your first choice fails.

OpenRouter passes through provider pricing while pooling uptime for better reliability. Each model and provider has a different price per million tokens. You only pay for what you actually use—there's no subscription or monthly bill. Some models also offer free access with a limited number of daily requests.

OpenRouter offers a free trial (approximately $1 in credits) with paid access afterward. You add credits to your account and draw from this shared credit pool as you make requests.

OpenRouter implements the OpenAI API specification for /completions and /chat/completions endpoints, allowing you to use any model with the same request/response format. Additional endpoints like /api/v1/models are also available. The API is fully OpenAI compatible, meaning existing OpenAI code will work with just a URL endpoint change.

The API supports text and images. Images can be passed as URLs or base64 encoded images. PDF and other file types are coming soon.

Yes, streaming is supported through the stream: true parameter.

For each model on OpenRouter, latency (time to first token) and token throughput information is displayed for all providers. You can use this to estimate how long requests will take. If you want to optimize for throughput, you can use the :nitro variant to route to the fastest provider.

OpenRouter supports several dynamic variants that change how requests are routed or used, including :extended which means the model has longer than usual context length, :thinking where the model supports reasoning by default, :online where all requests will run a query to extract web results attached to the prompt, :nitro which routes to the fastest provider for optimized throughput, :floor which prioritizes the lowest price, :beta which is not moderated by OpenRouter, and :exacto which uses only OpenRouter-curated high-quality endpoints.

OpenRouter logs basic request metadata including timestamps, model names, and token counts. However, prompts and completions are not logged by default.

For account and billing questions, email support@openrouter.ai. You can also browse OpenRouter's comprehensive guides, documentation, help center, and FAQ to find answers.

OpenRouter provides a Quick Start Guide, technical documentation, and API references through their developer portal. You can also check the OpenRouter Discord for updates on service status.