← News

Models & toolsSources: huggingface.co · 16 Nov 2026

Hugging Face switches language model routing to OpenRouter

Requests to language models on Hugging Face now go through OpenRouter's router, which picks a provider per call and retries failures elsewhere.

In plain words

Hugging Face lets developers run open-weight models — models whose weights are publicly downloadable — through its Inference Providers service. It has replaced its own routing system with OpenRouter's, which sends each request to whichever provider currently offers the best mix of price, speed and reliability, and retries on another provider if one fails.

What was announced

Hugging Face announced that Inference Providers, its service for running open-weight models, now routes language model requests through OpenRouter instead of its own custom router. OpenRouter selects a provider for each call based on live measurements of price, speed and reliability, and retries on another provider if the first attempt fails. The companies say the router supports OpenAI's Chat Completions and Responses APIs and Anthropic's Messages API, so existing code keeps working. Users can still pin a provider with a suffix on the model id, or disable providers in settings.

The change has one breaking effect: because requests are billed to the Hugging Face account, custom provider API keys no longer work for language models, though keys for image and video providers still do. Hugging Face says pricing is unchanged, with no markup added. OpenRouter reports that its rebuilt routing cut capacity errors by 46% and improved time to first token by 16.9% and throughput by 16.3% in its own measurements. The announcement also cites OpenRouter data showing open-weight models rose from 26.8% of tokens served in October 2025 to 71.7% at the end of September 2026, and notes that the best open-weight models still trail leading proprietary ones on hard benchmarks.

What it does not show

  • All figures, including the routing performance gains and open-weight usage share, come from Hugging Face's and OpenRouter's own announcements and are not independently confirmed.
  • The post describes a partnership's first step and planned future work; the full terms and long-term reliability of the routing are not established by this single announcement.

For security and IT teams

Access
Language model requests are billed to the Hugging Face account or organization, so custom provider API keys no longer work for those models; custom keys for image and video providers keep working.

Sources

Source-support review completed 17 Nov 2026. This story was drafted with AI assistance and checked against retained source passages. Automated checking is not independent verification.

Updates and corrections

No corrections so far.

First published 17 Nov 2026

New to the subject? AI from the Ground Up starts from the beginning.