Hugging Face Adds 3 New Inference Partners

Hyperbolic, Nebius AI Studio, and Novita have been added to the Hugging Face Hub as new serverless Inference providers. Following the existing Together AI, Sambanova, Replicate, fal, and Fireworks.ai, the ecosystem has expanded even further.

Through this partnership, the latest models such as DeepSeek-R1 and Flux.1 can now be run instantly across diverse infrastructure. Users can choose between two calling modes:

• Custom Key: Use each provider's own API key directly and pay that provider.
• Routed by HF: Use your Hugging Face token without a separate key, with costs billed to your Hugging Face account.

Integrated into the Python (huggingface_hub) and JavaScript (@huggingface/inference) SDKs, this offers the flexibility to easily switch inference engines by simply changing the provider argument value.