tools
Unsloth
Unsloth runs and fine-tunes AI models locally through a free, open-source desktop app, web UI, and code-based tools.
Unsloth lets developers run and train language, vision, image, video, audio, embedding, and diffusion models on local hardware. It supports fine-tuning methods such as LoRA, QLoRA, full fine-tuning, pretraining, GRPO, DPO, and reinforcement learning.
The product includes a desktop app, web UI, command-line tools, model export, dataset creation, web search, RAG, code execution, agent connections, and an OpenAI-compatible API. The standard version is free; Pro and Enterprise plans require contacting Unsloth for pricing.
Features
- Run and train language, vision, image, video, audio, embedding, and diffusion models
- Fine-tune with LoRA, QLoRA, full fine-tuning, pretraining, GRPO, DPO, and reinforcement learning
- Create datasets from PDFs, CSVs, DOCX files, and other sources
- Use local models with Claude Code, Codex, MCP, tool calling, and code execution
- Search the web, conduct deep research, and use retrieval-augmented generation
- Serve models through an OpenAI-compatible API and secure remote access
- Export models to GGUF, NVFP4, FP8, and other formats
- Open-source under Apache-2.0 and AGPL-3.0 licenses
Use cases
- Fine-tune a local language model for a private business dataset
- Run image and video generation without sending prompts to a hosted service
- Connect local models to coding agents such as Claude Code or Codex
- Build datasets from documents for training or retrieval-augmented generation
- Serve a locally hosted model through an OpenAI-compatible API
- Train and deploy models across NVIDIA, AMD, Intel, Apple, or CPU hardware
Pros
Cons
Pricing
- Starting price
- Free
- Pricing checked
- 2026-09-19
Free
Free
- Open-source
- Supports Mistral and Gemma
- Supports Llama 1, 2, and 3
- Supports 4 bit and 16 bit LoRA
unsloth Pro
Contact us
- 2.5x number of GPUs faster than FA2
- 20% less memory than OSS
- Enhanced MultiGPU support
- Up to 8 GPUs support
unsloth Enterprise
Contact us
- 32x number of GPUs faster than FA2
- Up to 30% accuracy
- 5x faster inference
- Full training
- Multi-node support
- Customer support