Run open-source AI models directly on your PC — fully offline, no subscription, no data leaving your machine.
Download for WindowsOpens LM Studio's official site · Windows 10/11, x64 or ARM (Snapdragon X Elite)
Publisher
LM Studio
Cost
Free
Runs
Fully offline
Format
GGUF / MLX
Once a model is downloaded, chat completely offline — nothing is sent to a server.
Search and download open-weight models like Llama, Mistral, and Qwen from inside the app.
Spin up an OpenAI-compatible endpoint on your machine for your own scripts and tools.
No account, no telemetry required — your prompts and documents never leave your device.
Click the download button above
It goes straight to lmstudio.ai, the developer's own site — not a third-party mirror.
Run the Windows installer
Standard .exe setup — no license key or sign-up needed to install.
Pick a model to download
Open the built-in "Discover" tab and grab a model sized for your RAM and GPU.
Start chatting locally
Load the model and chat in the app — entirely on your own hardware.
| OS | Windows 10 or 11 (x64 with AVX2, or ARM64) |
| RAM | 16 GB recommended (small models can run on less) |
| GPU | 4 GB+ VRAM recommended, not required |
| Disk | Varies by model — 4–40 GB per model file |