LM Studio0.4.28-2

Run open-weight large language models on your own computer with LM Studio, keeping prompts and conversations on the local machine. On Linux, this freeware Developer Tools desktop application executes models offline without relying on cloud processing, and it is also available for Windows and macOS.

Users can browse Hugging Face and obtain GGUF or MLX models directly, then interact with them through a built-in chat interface modeled on ChatGPT. It supports model families including Llama, Qwen, Mistral, Gemma, Phi, and DeepSeek, while organizing models, prompts, presets, and configurations in one workspace.

Document Chat applies retrieval-augmented generation to local files, allowing private offline conversations about their contents. An OpenAI-compatible API server exposes local models through REST endpoints, and the included CLI, Python SDK, TypeScript SDK, and REST tooling support application development. Headless mode runs the service without its graphical interface, making it suitable for servers and continuous-integration environments.

MCP integration connects MCP servers and external AI workflows to the local models. GPU offloading and memory controls let users adjust execution for computers with limited hardware, while local-network serving makes a model available to other devices on the same network. The current release is version 0.4.28-2.