Unsloth Studio
#9 in Local LLM RuntimesUnsloth · 2× · last seen Jul 29, 2026
27
Momentum
Unsloth Studio is a free, Open Source web interface (Beta) by Unsloth AI for locally training, running, and exporting Open-Source LLMs without coding knowledge. The software runs completely offline on your own machine (Mac, Windows, Linux) and supports Inference with GGUF and Safetensors models including Tool-Calling, web search, and an OpenAI-compatible API. It was launched on March 17, 2026 and complements the existing Unsloth core library (Apache 2.0) with a Studio UI component licensed under AGPL-3.0. Paid plans (Pro/Enterprise) primarily concern accelerated multi-GPU Training, not the Studio's core features.
Momentum trend
01.05.30.07.
Features
| Throughput/Latency | Training up to 2x faster with up to 70% less VRAM (no accuracy loss) |
| License | Dual license: core Unsloth Apache 2.0, Studio UI (studio/*) AGPL-3.0 |
| Platform | Windows, Linux, WSL and macOS; CPU operation possible (chat inference only) |
| Price | Free (open source); Pro/Enterprise tiers for multi-GPU acceleration available on request |
| Protocol Compatibility | OpenAI-compatible API as local endpoint; Model Context Protocol (MCP) server support |
| Release Date | March 17, 2026 (Beta launch) |
| Supported Models/Providers | 500+ models (incl. gpt-oss, Qwen3, Llama 4, Mistral, Gemma, Phi-4); GGUF, LoRA, safetensors; connects to OpenAI, Anthropic, vLLM, Ollama |