Hugging Face and OpenRouter are merging their systems so that running open-weight models requires no extra setup. Every major release on the Hugging Face Hub is now served through OpenRouter, which routes requests across more than 80 providers using live data on speed, price, and reliability. Your code remains unchanged, but language model requests now pass through OpenRouter to select the best provider for each call.
In this article
Use the tools you already have
The system supports OpenAI’s Chat Completions and Responses APIs, as well as Anthropic’s Messages API, regardless of the underlying provider. You can select a coding agent or SDK and a model, then follow the setup guides for Claude Code, Codex, OpenCode, Pi, or Hermes Agent. Your Hugging Face token also grants access to Storage Buckets, compute Jobs, and dedicated Inference Endpoints. PRO, Team, and Enterprise plans include monthly credits for compute and inference.
Adoption of open-weight models
Open-weight models accounted for 26.8% of tokens served in October 2025. By the end of September 2026, that figure had risen to 71.7%, following a peak of 77.5% in July 2026. Weekly volume for open-weight models grew more than 60-fold over this period, compared to nearly tenfold for closed models. DeepSeek V4.1 Flash 1017 is currently the most used model on OpenRouter, processing 59.2 trillion tokens.
Average costs for open-weight models are roughly four times lower than proprietary ones. Input tokens cost $0.28 per million and output tokens $1.62 per million, compared to $1.27 and $6.36 respectively for closed models. On the Artificial Analysis Intelligence Index, Kimi K3.1 Pro, the top open-weight model, scores 46 against 65 for Claude Fable 5.2. The cost per task for Kimi is $3.68 versus $26.54 for Claude.
Routing powered by OpenRouter
OpenRouter serves more than 800 trillion tokens monthly to over 10 million users across 600+ models from 80+ providers. The routing system now evaluates the best mix of price, speed, and reliability rather than selecting solely based on list price. If a provider slows down or fails, traffic moves to another source and the request retries on the next best provider within the same call.
In OpenRouter’s measurements, this new routing reduced capacity errors by 46%, improved time to first token by 16.9%, and increased throughput by 16.3%. You can still specify a policy or a specific provider using a suffix on the model id:
- zai-org/GLM-5.3: default selection based on price, speed, and reliability
- zai-org/GLM-5.3:fastest: highest throughput
- zai-org/GLM-5.3:cheapest: lowest price
- zai-org/GLM-5.3:novita: routes only to Novita
Routing is cache-aware; a conversation stays with the provider holding its prompt cache. To enforce this, send a stable UUID per conversation in the X-HF-Session-Id header.
Pricing
Pricing remains the same. You pay the provider’s rate on your Hugging Face bill, and Hugging Face adds no markup. PRO users receive $2 in monthly credits, while Team and Enterprise organizations get $2 per seat.
Image, video and audio generation
Image, video, and audio generation continue to work exactly as before through Hugging Face Inference Providers without passing through OpenRouter. You can use the same token and API to access tens of thousands of LoRAs.
Track your requests
On model pages, you can chat directly in the widget. Two new tabs show the infrastructure behind the model. The Providers list shows every available route with input and output prices, speed, and uptime over the last 24 hours. The Metrics tab displays the best price, throughput, time to first token, context length, and uptime among those options.
Your usage overview breaks down spend by the specific provider each request went to. In settings, you can disable any provider you do not wish to use; the router will never send requests there. These settings apply to organizations as well.
Breaking change: custom provider keys
Requests to language models now go through OpenRouter and are billed to your Hugging Face account or organization. You can no longer use your own API key for a provider with these models. You can still select a provider via a suffix on the model id or disable providers in your settings. Custom keys for image and video providers continue to work.
What’s next
This partnership marks the first step. Future plans include connecting your OpenRouter account to Hugging Face and routing more tasks through the system.



