Hugging Face is handing the job of routing its AI models to OpenRouter. The company said on November 16 that its Inference Providers service will now send language-model requests through OpenRouter's routing system, which picks the best provider for each call based on live measurements of price, speed and reliability — and automatically retries on another provider if the first one fails.
One token, more than 80 providers
OpenRouter serves more than 800 trillion tokens a month to over 10 million users, across more than 600 models from over 80 providers. For developers, nothing changes on the surface: the same Hugging Face token and API keep working, and the router now also supports OpenAI's Chat Completions and Responses APIs and Anthropic's Messages API for every chat model. Users can still steer routing with a suffix on the model name — :fastest for throughput, :cheapest for price, or a named provider such as :novita — and can switch off any provider in settings.
OpenRouter recently rebuilt its routing engine to weigh the best mix of price, speed and reliability instead of just the lowest list price. In its own measurements, the rebuild cut capacity errors by 46 percent, improved time to first token by 16.9 percent and lifted throughput by 16.3 percent. Image, video and audio generation are untouched: they still run directly through Inference Providers, and custom keys for those providers keep working.
The breaking change: your own provider keys stop working
The switch comes with one hard break. Requests to language models are now billed to the user's Hugging Face account or organisation, so developers can no longer use their own API key for a provider with these models. Anyone who previously paid a provider directly and passed their own key through Hugging Face will now be billed through Hugging Face instead. Hugging Face says pricing does not change — users pay the provider's price with no markup on top — and PRO users get $2 of credits a month.
Two new tabs on model pages show the picture: Providers lists every provider a model can be routed to, with price, speed and uptime, and Metrics shows the best figures across them. A usage overview breaks spending down by the provider each request went to. Hugging Face says this is only the first step of the partnership: next comes connecting an OpenRouter account to Hugging Face, and routing more tasks through OpenRouter.
Open-weight models keep getting cheaper — and busier
The move lands as open-weight models take over OpenRouter's traffic. Their share rose from 26.8 percent of tokens served in October 2025 to 71.7 percent at the end of September 2026, after peaking at 77.5 percent in July, while weekly open-weight volume grew more than 60-fold. The most-used model on OpenRouter this week is an open-weight one, DeepSeek V4.1 Flash 1017, at 59.2 trillion tokens.
Open-weight models also cost about four times less on average: $0.28 per million input tokens and $1.62 per million output tokens, against $1.27 and $6.36 for closed models. The best open-weight models still trail the best proprietary ones on the hardest benchmarks — Kimi K3.1 Pro scores 46 against 65 for Claude Fable 5.2 on the Artificial Analysis Intelligence Index — but at about a seventh of the cost per task. The backdrop is corporate, too: Nvidia's roughly $13 billion acquisition of Hugging Face is still pending, and the US Justice Department is probing the deal. For developers, the message is simple: one token reaches every open-weight model on the Hub — but the keys stay with Hugging Face.



