Artificial intelligence / news brief
Hugging Face moves chat-model inference routing to OpenRouter
- Event
- Published

Hugging Face switched language-model calls in Inference Providers to OpenRouter, which selects providers using live price, speed and reliability data and retries failed requests. Existing Hugging Face tokens and code continue to work. The change removes custom provider keys for language models, while image, video and audio routing remains outside the integration.
Why it matters
The switch centralizes routing and failover for a major open-model hub, improving convenience but changing billing and reducing direct provider-key control.
Sources
- OpenRouter Now Powers Inference on Hugging Face (external site)OpenRouter · primary source ·
- Hugging Face and OpenRouter join forces to make open-weight models easy to run (external site)Hugging Face · coverage ·
- Hugging Face will route language models through OpenRouter (external site)Dataist · coverage ·