OpenRouter is a unified API gateway that provides access to over 200 large language models (LLMs) from providers like OpenAI, Anthropic, Google, Meta, and Mistral through a single API endpoint. It handles model routing, fallbacks, and billing, making it easier to build AI applications without managing multiple API keys and provider accounts.

OpenRouter charges per token, split into input (prompt) tokens and output (completion) tokens. Costs are expressed in USD per million tokens ($/1M tokens). Output tokens are typically 3–5× more expensive than input tokens because they require more computation. You pay only for what you use, with no monthly minimums or seat fees.

A token is roughly 4 characters or 0.75 words in English text. So 1,000 tokens ≈ 750 words, or about 1.5 pages of text. Tokenisation varies by model — code, non-English languages, and special characters may tokenise differently. As a rule of thumb, 1 million tokens ≈ 750,000 words or 1,500 pages.

As of 2026, models like Llama 3.1 8B, Gemini Flash 1.5 8B, and Mistral 7B are among the most affordable on OpenRouter, with costs well under $0.10 per million tokens. Free-tier models with rate limits are also available. The cheapest suitable model depends on your quality requirements — smaller models are cheaper but may produce lower quality output for complex tasks.

Multiply your estimated daily requests by your average tokens per request (input + output), then multiply by the model’s per-token price. For example: 1,000 requests/day × 800 tokens each × 30 days = 24 million tokens/month. At $3/1M input + $15/1M output that equals roughly $180/month for typical input/output splits. This calculator does all that arithmetic for you automatically.

OpenRouter typically adds a small margin over the underlying provider’s prices to cover routing infrastructure. However, OpenRouter often negotiates volume discounts and passes savings on — meaning their effective price is sometimes equal to or lower than going directly to the provider, especially for smaller usage volumes. Always compare current prices at openrouter.ai/models for the latest rates.

Context length is the maximum number of tokens a model can process in a single request (input + output combined). Models with longer context windows (e.g., 128K or 1M tokens) cost more per token because they require more GPU memory. If your use case involves long documents or chat histories, you need a model with sufficient context length — but you will pay for every token in that context window on each call.

Yes. OpenRouter offers several free models including Llama 3.1 8B Instruct (free), Mistral 7B (free), and others — subject to rate limits. Free models are great for development and low-volume use cases. For production workloads with higher throughput requirements, paid models offer higher rate limits and guaranteed availability.