Skip to content- Sources: primary
- Summary: Hugging Face replaced its own router with OpenRouter's for language models, which picks a provider by live price, speed and reliability, retries failed calls on another provider and keeps conversations on the provider holding their prompt cache. Model ids accept suffixes such as :fastest, :cheapest or a provider name, and the API supports OpenAI Chat Completions and Responses and Anthropic Messages. As a breaking change, custom provider API keys no longer work for language models, while image, video and audio generation are unchanged.
- Why it matters: Anyone calling open-weight models through Hugging Face with their own provider keys must switch to Hugging Face billing or another path, and gains automatic failover.