Hugging Face routes language-model requests through OpenRouter


The routing change covers 160 models from 38 labs and lets developers keep using their Hugging Face code while OpenRouter selects a provider.
Hugging Face switched its Inference Providers service to OpenRouter’s routing system on 16 November 2026, handing it the job of selecting the third-party companies that run language-model requests. The integration covers 160 models from 38 labs: every model OpenRouter serves that has a page on Hugging Face Hub.
Previously, Hugging Face’s own system selected a provider and returned an error if that provider failed. OpenRouter says its system weighs live price, speed and reliability and can retry a failed request with another provider. Developers can keep using their existing Hugging Face token and code.
Hugging Face says chat, vision and agent requests go through OpenRouter. Image, video and audio generation remain on Hugging Face’s existing infrastructure.
Developers can prioritize speed or cost, choose a specific provider, compare available providers and turn off those they do not want to use. Organizations can apply the same restrictions to requests billed to them.
Language-model requests are billed to the user’s Hugging Face account or organization; users can no longer pay for those requests with their own provider API keys. Hugging Face says it charges the provider’s price without adding a markup. Custom provider keys still work for image and video requests.
The integration also works in the other direction: OpenRouter users can send request records to their own Hugging Face Storage Bucket for evaluation or fine-tuning, the process of further training a model. The records include inputs, outputs, token counts and cost, but leave out images and other media.
Hugging Face plans to let users link OpenRouter accounts and to route more tasks through OpenRouter. It has not given a timeline for either change, The New Stack said.
Selected customers can order the two-seat vehicle for personal use, but it cannot operate without a paid subscription. Tesla estimates delivery in Austin, Texas, in December 2026.
2 min

The Alliance for Creativity and Entertainment says the service’s operators agreed to transfer 16 domains to the coalition as part of the settlement.
1 min

The service connects compatible devices directly to satellites beyond cellular coverage. Iridium says it can provide coverage worldwide, though local availability depends on rules and providers.
2 min