FineWeb
FineWeb is Hugging Face's open 15-trillion-token dataset of cleaned, deduplicated Common Crawl web text, with a curated FineWeb-Edu subset. It is a high-quality, permissively licensed corpus for training modern LLMs.
Gradio
Gradio is an open-source Python library for quickly building demos and web applications for ML models, APIs, or any Python function. No JavaScript, CSS, or web hosting experience needed. Reached 1 million monthly developers in 2025. Features server-side rendering (v5.0) for faster page loads, sophisticated queuing for GPU-intensive workloads, streaming with Python generators, and seamless HuggingFace Spaces deployment. Deploy to HuggingFace for free with special Model Hub integration. Over 41,000 GitHub stars. Perfect for rapid ML prototyping and sharing.
Hugging Face Hub
The Hugging Face Hub is the central platform for sharing and discovering open models, datasets, and Spaces (demos), with versioning, an inference API, and libraries that make loading and deploying AI artifacts trivial.
Hugging Face Inference Providers
Inference Providers is Hugging Face's hosted routing layer: one Hugging Face token gives "access to hundreds of machine learning models, powered by world-class inference providers", including Baseten, Cerebras, Cohere, DeepInfra, Fal AI, Featherless AI, Fireworks, Groq, HF Inference, Novita, Nscale, OVHcloud, Public AI, Replicate, Scaleway, Together, WaveSpeedAI and Z.ai. It is reachable through the huggingface_hub and @huggingface/inference clients or as a drop-in OpenAI replacement at the base URL https://router.huggingface.co/v1, with provider selection policies (:fastest, :cheapest, :preferred) and automatic failover. Hugging Face charges "the same rates as the provider, with no additional fees", and every account gets monthly credits ($0.10 for free users, $2.00 for PRO, $2.00 per seat for Team or Enterprise). Two limits are worth knowing before adopting it: the OpenAI-compatible endpoint is "currently available for chat completion tasks only" (image, embedding and speech work must use the Hugging Face clients), and provider capability is uneven, with feature extraction, text-to-video and speech-to-text each supported by only a handful of the partners; using a custom provider key also means monthly credits do not apply and the provider bills you directly.
PEFT
PEFT is Hugging Face's library for parameter-efficient fine-tuning methods like LoRA, QLoRA, and prompt tuning, letting you adapt large models by training a tiny fraction of parameters on modest hardware.
TRL
TRL is Hugging Face's library for post-training transformer models with supervised fine-tuning, reward modeling, and preference optimization (DPO, GRPO, PPO), integrated with the Transformers and PEFT ecosystem.
Text Generation Inference
DISCONTINUED. Text Generation Inference (TGI) is Hugging Face's production-grade toolkit for serving open LLMs, with tensor parallelism, continuous batching, quantization, and an OpenAI-compatible API for high-throughput deployment. Hugging Face archived the repository on March 21, 2026 and placed the project in maintenance mode, accepting only minor bug fixes and documentation changes.
Transformers
Transformers is Hugging Face's model-definition framework for state-of-the-art machine learning in text, vision, audio, and multimodal models for both inference and training. Supports 300+ model architectures with 3 new architectures added weekly. Over 1M+ model checkpoints on Hugging Face Hub. Transformers v5 (2025) narrows to PyTorch as primary backend for deeper optimization, introduces 'transformers serve' for OpenAI-compatible API deployment, and enhances inference with streamlined APIs, continuous batching, and paged attention. Over 1.2 billion total installs with 3 million daily installations.
smolagents
smolagents is Hugging Face's minimalist agent library where agents write their actions as Python code. Small, hackable core with support for any LLM, tool calling, and sandboxed code execution.