Hugging Face | The AI community building the future.
huggingface.co ↗Hugging Face is a platform for hosting, discovering, and collaborating on machine learning models, datasets, and applications. The site provides access to millions of public resources and lists open-source tooling like Transformers, Diffusers, and Tokenizers. You can deploy models on inference endpoints or access paid compute and enterprise solutions with access controls.
Overview
Hugging Face is a collaboration platform for the machine learning community, described as "the AI community building the future." Its mission is to advance and democratize artificial intelligence through open source and open science. The platform hosts over 2 million models, 1 million applications (Spaces), and 500k+ datasets.
What it is / What it does
- Host and collaborate on unlimited public models, datasets, and applications
- Browse and discover ML models across modalities: text, image, video, audio, and 3D
- Spaces: deploy and run ML applications, with options for CPU or GPU acceleration
- Inference Providers: access 45,000+ models from multiple AI providers through a single unified API with no service fees
- Inference Endpoints: deploy models on optimized compute, starting at $0.60/hour for GPU
- Team and Enterprise plans with SSO, audit logs, resource groups, private datasets viewer, priority support, and dedicated regions, starting at $20/user/month
- Build and share an ML profile/portfolio
Stack / Implementation
The page lists the following open source libraries maintained by Hugging Face:
- Transformers — state-of-the-art AI models for PyTorch
- Diffusers — state-of-the-art diffusion models in PyTorch
- Safetensors — safe storage and distribution of neural network weights
- Hub Python Library — Python client to interact with the Hugging Face Hub
- Tokenizers — fast tokenizers optimized for research and production
- TRL — train transformer language models with reinforcement learning
- Transformers.js — state-of-the-art ML running directly in the browser
- smolagents — Python library to build agents
- PEFT — parameter-efficient finetuning for large language models
- Datasets — access and share datasets for ML tasks
- Text Generation Inference (TGI) — serve language models with an optimized toolkit
- Accelerate — train PyTorch models with multi-GPU, TPU, and mixed precision
Author / Source
Hugging Face (organization). GitHub: linked from the page at github.com/huggingface.
Notes
Over 50,000 organizations are listed as users, including AI2, Meta, Amazon, Google, Intel, Microsoft, Grammarly, and Writer. The page also links to HuggingChat, documentation, a blog, a forum, service status, and a changelog.