we_are_coded.by CODE · The world, decoded
БГ

Hugging Face · Infrastructure

Hugging Face switches its own router for OpenRouter's for open-weight models

Hugging Face announced on 16 November that Inference Providers moves from its own router to OpenRouter's. A failed request is now retried on another provider, and your own provider API key no longer works with the language models.

Source
Hugging Face
Published
The event
Checked on
Responsible editor
Cvetelin Ivanov
How we work
MethodCorrections

Until now, if a provider broke, you got its error. Now a failed try is retried on the next best provider, without you touching your code.

Hugging Face switched the router behind Inference Providers: the old one was its own. From 16 November, requests to language models go through OpenRouter's router, which picks a provider by live price, speed and reliability, and your code stays the same.

Fact

The facts: on 16 November 2026 Hugging Face announced it is switching Inference Providers from its own custom router to OpenRouter's. Requests to language models now go through OpenRouter, which picks the best provider for each call by price, speed and reliability and retries on another provider if the first one fails. The router supports OpenAI's Chat Completions and Responses APIs and Anthropic's Messages API for every chat model, whatever the provider behind it. According to OpenRouter, the service handles more than 800 trillion tokens a month for over 10 million users, across 600+ models from 80+ providers. According to OpenRouter, the share of open-weight models went from 26.8% of the tokens served in October 2025 to 71.7% at the end of September 2026, after peaking at 77.5% in July; weekly open-weight volume grew more than 60-fold, against almost tenfold for closed models. The most used model on OpenRouter is an open-weight one, DeepSeek V4.1 Flash 1017, with 59.2 trillion tokens. According to OpenRouter, open-weight models 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. On the Artificial Analysis Intelligence Index, the best open-weight model, Kimi K3.1 Pro, scores 46 against 65 for Claude Fable 5.2, at about a seventh of the cost per task ($3.68 against $26.54). In OpenRouter's measurements, the new routing cut capacity errors by 46% and improved time to first token by 16.9% and throughput by 16.3%. Pricing does not change: you pay the provider's price on your Hugging Face bill, with no markup. PRO users get $2 of credits every month, and Team and Enterprise organizations get $2 per seat.

The numbers are OpenRouter's, not Hugging Face's. The share of open-weight models there went from 26.8% to 71.7%. That is not a share of the whole world, but of the traffic passing through one router. Still enough to see why Hugging Face is not writing its own.

The cheapest no longer wins by default. The best mix of price, speed and reliability wins.

The breaking change is your key. Your own provider API key no longer works with these models: requests go through OpenRouter and are billed to your Hugging Face account. You can still pick a provider with a suffix on the model id (:fastest, :cheapest, :novita) or turn a provider off in your settings, and custom keys for image and video providers keep working.

Look at the Providers tab on a model page to see who can answer, and turn off the providers you don't want. The router will no longer send your requests there.

The visual is generated code-art, no third-party images.