unsloth.ai/docs/new/studio

Unsloth Studio is an open-source, no-code web interface for training, running, and exporting machine learning models entirely on your local machine — no cloud required. It supports 500+ models with optimised performance (2x faster training, 70% less VRAM) and works across Mac, Windows, and Linux. Designed for developers and researchers who want full control over fine-tuning and deployment without cloud costs or data privacy concerns. Key Features: - Local model execution — run GGUF and safetensor models offline on your own hardware - 500+ supported models — text, vision, text-to-speech, audio, and embedding models - Data Recipes — transform unstructured documents (PDF, CSV, JSON) into training datasets via visual workflows - Code execution — sandboxed Bash and Python environments built in - Model comparison — side-by-side inference testing between different models - Training observability — real-time monitoring of loss, gradients, and GPU utilisation - Export options — save fine-tuned models as safetensors or GGUF formats - 100% offline — no data leaves your machine Currently in beta. Dual-licensed under Apache 2.0 (core) and AGPL-3.0 (UI). Free and open-source.