OpenRouter and Hugging Face switched on a new integration on November 16, 2026. Language-model requests made through Hugging Face Inference Providers now run through OpenRouter's router, which selects a provider using live price, speed and reliability measurements and retries another provider when a request fails. Existing Hugging Face tokens and application code continue to work.
According to OpenRouter, 160 models from 38 labs are available through the integration at launch. Users can prioritize the fastest or cheapest provider with a model-ID suffix, select a specific provider, or disable providers they do not want to use. Hugging Face says pricing remains the provider's price on the Hugging Face bill, without an added markup.
There is one breaking change: custom API keys for individual providers no longer work for language-model requests. Those requests are instead billed to the Hugging Face account or organization. Custom keys for image and video providers remain supported, and image, video and audio generation do not use the new OpenRouter route.
OpenRouter can also send request traces to a Hugging Face Storage Bucket owned by the user through its Broadcast beta. The JSON files contain input messages, model output, token counts and cost; images and other media are excluded.