One Hugging Face token to run open models for chat, images, video, speech and embeddings, served by partner providers such as Replicate, Together and Groq. Call it from Hugging Face's Python and JavaScript clients or an OpenAI-compatible endpoint.
Free users buy credits to use it; PRO gets $2.00 a month, Team and Enterprise $2.00 per seat.
Hugging Face passes provider costs through with no markup.
Logs kept up to 30 days for debugging, with no user data or tokens stored.
The @huggingface/inference client is MIT licensed; the routed inference service is hosted.
huggingface_hub InferenceClient for Python and @huggingface/inference for JavaScript; OpenAI clients work for chat.
Nothing new in the last 30 days. For older changes, see the official changelog.
Wirestack