Hugging Face Integrates Fireworks.ai as an Inference Provider

Hugging Face has integrated Fireworks.ai as a supported Inference Provider on the Hub. This integration allows users to access blazing-fast serverless inference for a wide range of models directly on model pages and across the Hugging Face ecosystem of libraries and tools.

Supported Models

Fireworks.ai provides serverless inference for numerous models on the Hugging Face Hub, including:

  • DeepSeek-R1 (deepseek-ai/DeepSeek-R1)
  • DeepSeek-V3 (deepseek-ai/DeepSeek-V3)
  • Mistral-Small-24B-Instruct-2501 (mistralai/Mistral-Small-24B-Instruct-2501)
  • Qwen2.5-Coder-32B-Instruct (Qwen/Qwen2.5-Coder-32B-Instruct)
  • Llama-3.2-90B-Vision-Instruct (meta-llama/Llama-3.2-90B-Vision-Instruct)

Implementation and Access Methods

Users can access Fireworks.ai inference capabilities through three primary interfaces:

Website User Interface

Users can search for and run inference on all models supported by Fireworks.ai directly within the Hugging Face website UI.

Client SDKs

Fireworks.ai can be integrated into applications using the following SDKs:

  • Python (huggingface_hub): By installing the library from source and defining the provider="fireworks-ai" parameter in the InferenceClient, users can route requests to Fireworks.ai. Authentication can be handled via a Hugging Face token or a Fireworks.ai API key.
  • JavaScript (@huggingface/inference): Users can specify the provider: "fireworks-ai" parameter within the chatCompletion method to utilize the provider.

HTTP Calls

Direct HTTP requests can be made via cURL to the Hugging Face router endpoint (https://router.huggingface.co/fireworks-ai/v1/chat/completions), specifying the model and authentication bearer token.

Billing and Credits

Billing for Fireworks.ai inference is handled based on the authentication method used:

  • Direct Requests: When using a Fireworks.ai API key, users are billed directly through their Fireworks account.
  • Routed Requests: When authenticating via the Hugging Face Hub, users pay standard Fireworks API rates with no additional markup from Hugging Face.

Additionally, Hugging Face PRO users receive $2 worth of Inference credits every month, which can be applied across different providers.

Sources