Hugging Face switches Inference Providers to OpenRouter routing

·2026.11.16 09:00

Key point

The switch enables automatic provider selection based on live price, speed, and reliability, with automatic retries on failure.

1 / 8

Details

  • Hugging Face Inference Providers for language models now route requests through OpenRouter instead of a custom router.
  • OpenRouter selects the best provider per request based on live price, speed, and reliability, retrying on a different provider if the first fails.
  • Open-weight models account for 71.7% of tokens served on OpenRouter as of September 2026, up from 26.8% in October 2025.
  • Open-weight models cost approximately four times less on average, with $0.28 per million input tokens and $1.62 per million output tokens, versus $1.27 and $6.36 for closed models.
  • The new routing system cut capacity errors by 46% and improved time to first token by 16.9%.
  • Custom provider API keys are no longer supported for language models; requests are billed to the Hugging Face account.
  • Image, video, and audio generation continue to operate directly through Hugging Face without OpenRouter routing.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.