Unsloth

Python package for code-based training and inference.

Unsloth Notebooks is a collection of free fine-tuning notebooks for training models, powered by free GPU compute, supporting various models and tasks.

Unsloth Start: For the complete documentation index, see llms.txt .

Next-gen quantization for LLMs, preserving quality at smaller sizes.

Unsloth is an open-source library for fine-tuning LLMs with 2x faster training and 70% less memory usage.

Native desktop app for macOS, Windows, and Linux.

Unsloth Studio is an open-source web UI for training and running models with no-code and observability.

Directory of Unsloth Dynamic GGUF, 4-bit, and NVFP4 models on Hugging Face, including Qwen, DeepSeek, Llama, Gemma, and more.

Run local LLMs with an OpenAI- and Anthropic-compatible API endpoint, featuring self-healing tool calling, code execution, and web search.

Comprehensive guide to Reinforcement Learning (RL) with Unsloth, covering GRPO, reward functions, and training reasoning models.