Hugging Face language model inference routing switched to OpenRouter
Understand this story
Hugging Face has switched language model requests for Inference Providers from its self-built router to OpenRouter, which selects providers based on real-time price, speed, and reliability, and automatically retries on failure. OpenRouter said existing Hugging Face tokens and code do not need to be changed. Hugging Face said users still pay provider prices on their bills, with no markup by the platform; however, language model requests are now billed through OpenRouter to the Hugging Face account or organization, and users can no longer use their own provider API keys for these models. Image, video, and audio generation are unaffected and are still completed directly through Hugging Face Inference Providers, not through OpenRouter. PRO users receive $2 in credit per month, and Team and Enterprise organizations receive $2 per seat.
AI overview based on source reports · Updated 42 min ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Hugging Face BlogFeaturedHugging Face and OpenRouter join forces to make open-weight models easy to run
Hugging Face has switched its Inference Providers router to OpenRouter's, so requests to language models on the Hub are routed across 80+ providers based on live price, speed and reliability, with failed calls retried on another provider while code and the Hugging Face token stay the same. The router supports OpenAI's Chat Completions and Responses APIs and Anthropic's Messages API, and model ids can carry suffixes like :fastest, :cheapest or a provider name; image, video and audio generation still run directly through Hugging Face. Hugging Face 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 a breaking change: custom provider API keys no longer work for language models.
- OpenRouter AnnouncementsOpenRouter Now Powers Inference on Hugging Face
OpenRouter's provider routing now powers Hugging Face Inference Providers, selecting providers per request based on live price, speed, and reliability with automatic failover, while existing Hugging Face tokens and code keep working. OpenRouter can also send request traces as JSON files to a Hugging Face Storage Bucket you own, and 160 models from 38 labs are available this way today.
Activity for this story
Not enough continuous observations to show a trend yet.