AI Briefing
KO

Fireworks.ai added as an HF Inference Provider

·2025.02.14 09:00

Key point

Fireworks.ai has been added as a serverless inference provider on the Hugging Face Hub, supporting major LLMs.

1 / 2

Details

Fireworks.ai is now officially supported as an Inference Provider on the Hugging Face Hub. This allows developers to use Fireworks.ai's high-speed serverless inference capabilities directly from model pages within the Hugging Face ecosystem.

Key supported models include the following:

  • DeepSeek-R1, DeepSeek-V3
  • Mistral-Small-24B-Instruct-2501
  • Qwen2.5-Coder-32B-Instruct
  • Llama-3.2-90B-Vision-Instruct

Users can easily call these not only through the Hugging Face website UI, but also via the Python (huggingface_hub) and JavaScript (@huggingface/inference) SDKs, as well as HTTP/cURL.

For billing, Fireworks's standard rates apply whether you use your Fireworks.ai API key directly or route through Hugging Face. In addition, Hugging Face PRO users receive $2 worth of inference credits per month that can be used across various providers.

huggingfacellm-servingfireworks-aiinference-providerserverless-inference
Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.

Related

HuggingFace BlogHugging Face Integrates BasetenPyTorchKRvLLM Releases 'Decision 1.0', AI Models Specialized for Decision-MakingTLDR AIFireworks Releases Ember-1: Matches Kimi K3's Performance with 40% Fewer TokensRedditHemmingway-1, a 27B Creative Writing Model Based on Qwen3.8, ReleasedPyTorchKRRecipe Released for Serving 320B GLM-5.3-Flash on Free Kaggle TPUsAWS Tech Blog (Korea)Optimizing Gemma 4 Serving with vLLM on Amazon EKS: Cold Start Reduced from 428s to 226s