The Hugging Face OpenRouter integration is live: Hugging Face announced on November 16, 2026 that Inference Providers, its serverless inference service, now uses OpenRouter's router for language models. Each request is now sent to the provider with the best mix of price, speed and reliability at that moment, measured in real time — and if a provider fails, the attempt is retried on another provider inside the same call.
What changes in routing
Before the change, Inference Providers picked a single provider per request and returned its error when something failed. With OpenRouter routing, traffic is spread across more than 80 providers, with automatic failover: when a provider slows down or starts failing, requests move to others until it recovers. According to OpenRouter's measurements of the routing rebuilt in October, the new logic cut capacity errors by 46%, sped up time to first token by 16.9% and raised throughput by 16.3%, at 6.0% higher cost per request.
Which models are covered
The change covers 160 models from 38 labs — every model OpenRouter serves that has a page on the Hub — including DeepSeek V4.1 Flash, Kimi K3.1 Pro and K3, GLM 5.3 and GLM 5.3 Flash, Qwen3.8 27B, MiniMax M3.1 Flash, MiMo-V2.6-Flash, Nemotron 3 Ultra and gpt-oss-120b. The router supports OpenAI's Chat Completions and Responses APIs and Anthropic's Messages API, and your Hugging Face code and token stay the same. To prioritize something, you can add suffixes to the model ID, such as :fastest, :cheapest or a provider name, for example zai-org/GLM-5.3:fastest.
Open-weight models on the rise
The backdrop is the rise of open-weight models: on OpenRouter, they went from 26.8% of tokens served in October 2025 to 71.7% at the end of September 2026, peaking at 77.5% in July, with weekly volume growing more than 60-fold over the period. The most used model this week is an open-weight one, DeepSeek V4.1 Flash 1017, at 59.2 trillion tokens. On average, open-weight models cost about four times less: $0.28 per million input tokens and $1.62 per million output tokens, against $1.27 and $6.36 for closed models.
Warning: bring-your-own keys stop working
There is one breaking change: bring-your-own provider keys stop working for language models, because requests are now billed to your Hugging Face account. You can still pick a provider with a suffix on the model ID or turn providers off in settings. Custom keys keep working for image and video providers, and image, video and audio generation continue through the direct Inference Providers path, without going through OpenRouter. Pricing does not change: you pay the provider's price on your Hugging Face bill, with no added markup.
What's next
The integration also runs the other way: OpenRouter can send a trace of each request to a Hugging Face Storage Bucket you own — in beta — with messages, output, token counts and cost, raw material for building evaluation or fine-tuning sets. Model pages gained Providers tabs, with each provider's price, speed and uptime, and Metrics tabs. The change comes weeks after Stripe completed its acquisition of OpenRouter, as Geekiki reported. According to the companies, this is only the first step of the partnership: next comes connecting your OpenRouter account to Hugging Face and routing more task types through OpenRouter.
Sources:
Image: Hugging Face chart showing the share of open-weight models in tokens served by OpenRouter, from 26.8% in October 2025 to 71.7% in September 2026. Source: Hugging Face blog.
This article was produced with the assistance of artificial intelligence, cross-checked against public posts and internet sources. Consult the original sources for more details.
Por Muse