GitHub radar
Unsloth Studio: Run and Fine-Tune LLMs on Your Machine
The first desktop app to run and fine-tune language, diffusion, voice, and embedding models locally — with a web UI, drag-and-drop dataset builder, and no cloud required.
Unsloth is described by its authors as "the first desktop app to run and train models." Its web UI, called Unsloth Studio (launched with the command `unsloth studio`), lets you browse a model catalog, download, and run LLMs, diffusion, TTS, and embedding models locally. A drag-and-drop dataset builder accepts PDFs, CSVs, and DOCX files for fine-tuning without any code. Runs on Windows, macOS, and Linux with NVIDIA, AMD, Intel GPUs, CPUs, and Vulkan. You can optionally connect to ChatGPT or other cloud providers. Benchmark tables cover models from Google (Gemma), OpenAI (gpt-oss), Alibaba (Qwen), and Meta (Llama). Authors benchmark fine-tuning at 2× faster with 70% less VRAM versus standard pipelines, and MoE model preparation at 12× faster with 35% less VRAM. The project has 75 000+ GitHub stars.
Why a vibe-coder should care
If you want to run a language model without a cloud subscription, Unsloth Studio gives you a working local interface. If you want to fine-tune a model on your own data — company documents, customer Q&A, writing examples — you can do it through a GUI that accepts PDFs and spreadsheets, with no coding.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Install Unsloth from https://github.com/unslothai/unsloth — follow the README, run `unsloth studio`, open the web UI in the browser, and show me the list of available models. Ask if you need permissions or additional steps.
Запуск небольших моделей — обычный ноутбук, видеокарта необязательна. Для крупных моделей и дообучения нужна GPU: NVIDIA, AMD, Intel или Mac на M-чипе.
Open on GitHub▌ More finds