Hugging Face switches model routing to OpenRouter

Hugging Face is switching its Inference Providers for language models to OpenRouter, allowing users to call eligible open-weight models through the same API while OpenRouter selects a provider for each request. The partnership is designed to improve access, reliability and cost visibility without requiring code changes.
- Hugging Face is switching language model routing to OpenRouter.
- OpenRouter selects providers using live price, speed and reliability data.
- Users can access eligible open-weight models with the same Hugging Face API.
- Pricing stays unchanged, with no Hugging Face markup.
- Custom provider keys no longer work for language model requests.
One API for open-weight models
Every major open-weight release is already published on the Hugging Face Hub, while OpenRouter serves models at scale across 80+ providers. Under the new arrangement, every open-weight model on the Hub that a provider serves can be called with a Hugging Face token through the same API.
Requests to language models now go through OpenRouter. The router chooses a provider using live measurements of price, speed and reliability, and retries the request with another provider if the first one fails. Hugging Face says that users' code stays the same.
The router supports OpenAI's Chat Completions and Responses APIs and Anthropic's Messages API for every chat model, regardless of the provider behind it. Hugging Face also provides setup guides for Claude Code, Codex, OpenCode, Pi and Hermes Agent.
Open-weight usage is growing
According to figures cited by Hugging Face from OpenRouter, open-weight models accounted for 26.8% of tokens served in October 2025 and 71.7% at the end of September 2026, after reaching 77.5% in July. Weekly open-weight volume grew more than 60-fold over the same period, compared with almost tenfold for closed models.
DeepSeek V4.1 Flash 1017 is currently the most used model on OpenRouter, with 59.2 trillion tokens. Hugging Face says open-weight models cost about four times less on average: $0.28 per million input tokens and $1.62 per million output tokens, compared with $1.27 and $6.36 for closed models.
The source also points to a performance trade-off. On the Artificial Analysis Intelligence Index, Kimi K3.1 Pro, described as the best open-weight model, scores 46, compared with 65 for Claude Fable 5.2. The cited cost per task is $3.68 for Kimi K3.1 Pro and $26.54 for Claude Fable 5.2.
Routing, pricing and controls
OpenRouter serves more than 800 trillion tokens a month to over 10 million users across 600+ models from 80+ providers. Its recently rebuilt routing system looks beyond the lowest list price and selects the provider with the best mix of price, speed and reliability. In OpenRouter's measurements, the change cut capacity errors by 46%, improved time to first token by 16.9% and increased throughput by 16.3%.
Users can still select a routing policy or a specific provider by adding a suffix to the model ID, such as :fastest, :cheapest or :novita. They can also turn off providers in their settings. Hugging Face says routing is cache-aware, keeping a conversation with the provider holding its prompt cache.
Pricing does not change. Users pay the provider's price on their Hugging Face bill, with no markup from Hugging Face. PRO users receive $2 of credits every month, while Team and Enterprise organizations receive $2 per seat. Image, video and audio generation continue to run directly through Hugging Face Inference Providers.
The change also means that language model requests are billed to a Hugging Face account or organization, so users can no longer use their own provider API key for these models. Custom keys for image and video providers continue to work. Hugging Face says the next step is to connect OpenRouter accounts to Hugging Face and route more tasks through OpenRouter.
Hugging Face: Hugging Face and OpenRouter join forces to make open-weight models easy to run


