Inference Providers documentation
Providers
Providers
Inference Providers sends each request to a provider serving the model. You pay that provider’s price with no markup, and you can turn off the providers listed on this page in your settings.
Language models
Chat, vision and agent requests go through OpenRouter, which routes each one to a provider serving the model, including the ones below. See Routing and Providers to choose how a provider is picked. To send a request to one provider only, append the suffix shown on its card to the model id, for example deepseek-ai/DeepSeek-V4.1-Flash:together.
Baseten
Cerebras
Cohere
CoreWeave
DeepInfra
DigitalOcean
Featherless AI
Fireworks
Liquid AI
Mistral
Moonshot AI
Nebius
Novita
Parasail
Prime Intellect
:primeintellect · On the HubReka
Tencent
Together AI
Zai
Which providers serve a language model
A model’s page on the Hub lists its providers. To compare their price, speed and context length, use the models page or GET https://huggingface.co/proxy/router.huggingface.co/v1/models (see Hub API).
Image, video, audio, embeddings and more
Other tasks go to the following providers through the Hugging Face SDKs. Each link lists the models a provider serves on the Hub. To pick a provider, pass the id shown on its card as provider. The list of providers is also available on each model’s page.