unslothai / unsloth
Unsloth: The First Desktop App to Run and Train AI Models
Unsloth：首款用于运行和训练 AI 模型的桌面应用
Unsloth is the first desktop app to run and train models. Unsloth 是首款用于运行和训练 AI 模型的桌面应用。
⚡ Get started
⚡ 快速开始
Download the native Unsloth Desktop app for your operating system: 下载适用于您操作系统的原生 Unsloth 桌面应用：
- Windows: Download
- macOS: Download
- Linux / Ubuntu (deb): Download
- Linux (AppImage): Download
- Linux (Arm64): Download
Download from Unsloth or GitHub Releases. Or if you prefer to install manually: 您可以从 Unsloth 官网或 GitHub Releases 下载。如果您更喜欢手动安装：
- macOS, Linux, WSL:
curl -fsSL https://unsloth.ai/install.sh | sh - Windows:
irm https://unsloth.ai/install.ps1 | iex
Community: Discord | 𝕏 (Twitter) | Reddit
⭐ Features
⭐ 功能特性
Unsloth lets you run, train, and deploy AI models locally, with support for all types of models. Unsloth 让您能够在本地运行、训练和部署 AI 模型，并支持所有类型的模型。
Run & Build with AI 使用 AI 运行与构建
- Run and train LLMs, diffusion, embedding, audio models: Kimi K3, MiniMax-H3, Qwen3.8, Muse Glimmer, DeepSeek-V4, Gemma 4. 运行和训练大语言模型 (LLM)、扩散模型、嵌入模型和音频模型：包括 Kimi K3、MiniMax-H3、Qwen3.8、Muse Glimmer、DeepSeek-V4、Gemma 4 等。
- Agents & Tools: Use local models with Claude Code, Codex, and MCP, including tool calling and code execution. 智能体与工具：将本地模型与 Claude Code、Codex 和 MCP 结合使用，支持工具调用和代码执行。
- Search & RAG: Use private and unlimited web search, deep research, and RAG. 搜索与 RAG：使用私有且无限制的网络搜索、深度研究和检索增强生成 (RAG)。
- Image and video: Run and train image and video diffusion or multimodal models. 图像与视频：运行和训练图像及视频扩散模型或多模态模型。
- Hardware: Supports CPU, NVIDIA, AMD, Intel, macOS, and multi GPU setups. 硬件支持：支持 CPU、NVIDIA、AMD、Intel、macOS 以及多 GPU 配置。
- Remote Access: Access your local models remotely through secure Cloudflare HTTPS. 远程访问：通过安全的 Cloudflare HTTPS 远程访问您的本地模型。
Train & Deploy 训练与部署
- Fine-tuning: Train LLMs, diffusion, TTS, and embedding models 2× faster with 70% less VRAM. 微调：训练大语言模型、扩散模型、TTS 和嵌入模型，速度提升 2 倍，显存占用降低 70%。
- Complete support: Supports reinforcement learning, LoRA, QLoRA, full fine tuning, pretraining, RL, GRPO, DPO, and FP8. 全面支持：支持强化学习、LoRA、QLoRA、全参数微调、预训练、RL、GRPO、DPO 和 FP8。
- Export & Deploy: Export or Deploy models including GGUF, NVFP4, FP8 and more formats. 导出与部署：导出或部署模型，支持 GGUF、NVFP4、FP8 等多种格式。
- Datasets: Build datasets from PDFs, CSVs, DOCX files, and more with Data Recipes. 数据集：通过 Data Recipes 从 PDF、CSV、DOCX 文件等构建数据集。
- OpenAI Compatible API: Serve models through an OpenAI compatible API and also connect to cloud providers. 兼容 OpenAI 的 API：通过兼容 OpenAI 的 API 提供模型服务，并可连接至云服务提供商。
🚀 Unsloth Start
🚀 Unsloth Start
Unsloth Start connects Claude Code, Codex and other agents to local models with one command. Unsloth Start 只需一条命令即可将 Claude Code、Codex 和其他智能体连接到本地模型。
Start Unsloth, load a model, open your project folder, then run: 启动 Unsloth，加载模型，打开项目文件夹，然后运行：
unsloth start claude
Replace claude with any supported agent:
将 claude 替换为任何受支持的智能体：
| Agent | Command |
|---|---|
| Claude Code | unsloth start claude |
| OpenAI Codex | unsloth start codex |
| Hermes Agent | unsloth start hermes |
| OpenClaw | unsloth start openclaw |
| OpenCode | unsloth start opencode |
Claude Code, Codex and OpenCode can keep their current model and use Unsloth as a local subagent: Claude Code、Codex 和 OpenCode 可以保留其当前模型，并将 Unsloth 作为本地子智能体使用：
unsloth start claude --as-subagent --model unsloth/model-GGUF:quant
📥 Install
📥 安装
Unsloth can be used in three ways: Unsloth Desktop, the desktop app; Unsloth Studio, the web UI; or Unsloth Core, the code based version. Unsloth 有三种使用方式：Unsloth Desktop（桌面应用）、Unsloth Studio（Web UI）或 Unsloth Core（基于代码的版本）。
Unsloth Desktop (recommended) Unsloth Desktop（推荐）
The Tauri based desktop app is the easiest way to use Unsloth and needs no setup, so start here. 基于 Tauri 的桌面应用是使用 Unsloth 最简单的方式，无需配置，建议从这里开始。
(Platform links omitted for brevity, same as above)
Unsloth Studio (web UI) Unsloth Studio (Web UI)
Unsloth Studio (Beta) works on Windows, Linux, WSL and macOS. Unsloth Studio (Beta) 适用于 Windows、Linux、WSL 和 macOS。
- CPU: Supported for Chat and Data Recipes currently. CPU：目前支持聊天和 Data Recipes。
- NVIDIA: Training works on RTX 30/40/50, Blackwell, DGX Spark, Station and more. NVIDIA：支持在 RTX 30/40/50、Blackwell、DGX Spark、Station 等设备上进行训练。
- macOS: Training, MLX and GGUF inference are ALL supported. macOS：全面支持训练、MLX 和 GGUF 推理。
- AMD: Training, RL, chat and deployment work on Windows, WSL and Linux. AMD：支持在 Windows、WSL 和 Linux 上进行训练、强化学习、聊天和部署。
- Vulkan: GGUF inference is supported on compatible GPUs, including Intel GPUs. Vulkan：支持在兼容的 GPU（包括 Intel GPU）上进行 GGUF 推理。
- Multi-GPU: Available now, with a major upgrade on the way. 多 GPU：现已可用，重大升级即将推出。