Models.dev, an open-source registry of AI model specifications and pricing, shipped its public release on GitHub on May 26. The project consolidates context windows, modality support, rate limits, and per-token prices across the major model providers into a single dataset, accessible via API rather than only through a web interface.
The competitive position is specific. Artificial Analysis and OpenRouter both publish similar comparison data, but Artificial Analysis is a paid product with a heavy UI and OpenRouter is a routing layer where the comparison is incidental to the inference brokerage. The various awesome-llm and llm-pricing GitHub lists are static and quickly fall behind. Models.dev is the lightweight programmatic answer: a structured dataset designed to be queried from automated pipelines, model-selection logic, and procurement tooling.
For builders writing model-routing code, the practical value is the ability to make per-request decisions about which model to call based on current price and capability data without scraping vendor pages. Pull the spec, branch on context-window requirement, sort by price per million tokens, and you have a basic model router in a few lines of code. The same data is useful for cost forecasting and contract negotiation, where being able to ground a discussion in a current third-party reference matters more than the precise number.
Released on GitHub by the anomalyco organization on 2026-05-26.
