Posts

Showing posts with the label OpenRouter

Hugging Face Hands Model Routing to OpenRouter — and Your Own API Keys Stop Working

Image
Hugging Face switched the routing behind its Inference Providers to OpenRouter on 16 November 2026, with a sharp edge for developers: your own API keys for a model provider no longer work for language models. Requests now go through OpenRouter, which picks the best provider for each call based on live price, speed and reliability, and retries on another provider if the first one fails. Your code stays the same. Model routing is the job of choosing which company actually runs a model request. A single open-weight model can be hosted by several providers at different prices and speeds, so routing decides who gets each call. Under the old system, that decision was made once and a failure ended the request; under the new system, it is made live for every call, and a failure triggers a retry somewhere else. What changed on 16 November Inference Providers is the Hugging Face service that lets developers call open-weight models through one API instead of setting up hosting themselves. U...