Skip to content
NewTry RouterPlus with free creditsTry playground →
← All posts · News

Hugging Face Inference Providers now routes LLM calls via OpenRouter

Same token and code, provider failover added, and custom provider keys for language models are gone.

Hugging Face switched its Inference Providers service from its own router to OpenRouter's on November 16, 2026. Language model calls made with a Hugging Face token now go through OpenRouter, which picks a provider for each request and retries on another provider if the first one fails. Hugging Face says existing code keeps working.

What changed

  • Chat model requests on Inference Providers now run through OpenRouter's router, over OpenAI Chat Completions, OpenAI Responses or Anthropic Messages. Image, video and audio generation still run directly through Hugging Face, as before.
  • Breaking change: you can no longer use your own API key for a provider with language models. Those calls are billed to your Hugging Face account or organization. Custom keys for image and video providers keep working.
  • Routing is set with a suffix on the model ID: :fastest, :cheapest or a provider name such as :novita. With no suffix, OpenRouter weighs price, speed and reliability. A conversation stays on the provider holding its prompt cache; send a stable UUID in the X-HF-Session-Id header to make that explicit.
  • Pricing is unchanged: you pay the provider's price on your Hugging Face bill, with no Hugging Face markup. PRO users get $2 of credits a month; Team and Enterprise organizations get $2 per seat.
  • OpenRouter says 160 models from 38 labs are available this way. It can also send a trace of each request to a Hugging Face Storage Bucket you own, through its Broadcast feature (in beta).

What to do

If you call open-weight chat models through Hugging Face with your own provider key, those calls now bill to Hugging Face. To pin a provider or a policy, add a suffix to the model ID, as in Hugging Face's examples:

zai-org/GLM-5.3            # default routing
zai-org/GLM-5.3:cheapest   # lowest price
zai-org/GLM-5.3:novita     # only Novita

To keep billing on your own provider account, call that provider's API directly. Hugging Face lists connecting an OpenRouter account and routing more tasks through OpenRouter as next steps, with no dates.

GLM 5.3 on RouterPlus

RouterPlus serves open-weight models from several hosts behind one API key at api.routerplus.com/v1, billed from prepaid credits at listed model rates. GLM 5.3, the model in Hugging Face's examples, per million tokens from the RouterPlus catalog on November 16, 2026:

ProviderInputCached inputOutput
Wafer$0.20$0.18$1.80
DeepInfra$0.5625$0.125$2.50
Inference.net$0.90$0.15$3.00
Novita AI$0.91$0.169$2.86
Sail Research$0.98$0.18$3.08
Morph$1.19$0.20$6.00
RunInfra$1.30$0.13$4.20
Alpha.sh$1.40$0.26$4.40
Doubleword$1.40$0.28$4.40
Mistral AI$1.40$0.14$4.40
Nebius$1.40$0.26$4.40
StreamLake$1.40$0.26$4.40

RouterPlus bring your own key covers your own OpenAI, Anthropic or Azure keys and AWS Bedrock roles, with no RouterPlus fee on those calls. See the GLM 5.3 model page and all Z.ai models on RouterPlus.

Sources

  1. Hugging Face: Hugging Face and OpenRouter join forces to make open-weight models easy to run (November 16, 2026)
  2. OpenRouter: OpenRouter Now Powers Inference on Hugging Face (November 16, 2026)
  3. RouterPlus model catalog (November 16, 2026)
  4. RouterPlus docs: Bring your own key