# LM Studio
Cross-platform desktop app to discover, download, and run local [[Large Language Models (LLMs)]]. Runs entirely offline once models are downloaded.
## Supported Platforms
- **macOS** — Apple Silicon (M1–M4), macOS 14.0+
- **Windows** — x64 and ARM (Snapdragon X Elite)
- **Linux** — x64 and ARM64
## Supported Model Formats
- **GGUF** via `llama.cpp`
- **MLX** on Apple Silicon Macs
## Core Features
- Chat UI with document attachment (RAG) for offline analysis
- Model discovery and download from [[HuggingFace]] (keyword or URL search)
- OpenAI-compatible API (`/v1/chat/completions`, `/v1/embeddings`)
- Anthropic-compatible API (`/v1/messages`) — since v0.3.29
- Native LM Studio REST API with stateful chats and [[Model Context Protocol (MCP)]] support — since v0.4.0
- Tool calling / function calling
- Speculative decoding for faster generation
- Parallel request processing via continuous batching
- Per-model configuration defaults and shareable presets
- Headless deployment via `llmster` daemon (no GUI, for servers/VMs)
## Claude Code Integration
Since v0.4.1, LM Studio can serve as a backend for [[Claude Code]] via its Anthropic-compatible `/v1/messages` endpoint.
### Setup
1. Start the server: `lms server start --port 1234`
2. Set environment variables:
- `ANTHROPIC_BASE_URL=http://localhost:1234`
- `ANTHROPIC_AUTH_TOKEN=lmstudio`
3. Run: `claude --model openai/gpt-oss-20b` (or any loaded model)
For VS Code, set the same variables in `claudeCode.environmentVariables`.
### Notes
- Minimum context size: 25K tokens (more is better)
- Works with GGUF and MLX models
- Model capability for agentic tasks varies — not all models handle tool use well
## Mobile
See [[LM Studio Mobile App]] — the official iOS app (Locally) that uses your phone as a remote client for the models running on your desktop, via the end-to-end encrypted LM Link feature.
## References
- Official website: https://lmstudio.ai
- Documentation: https://lmstudio.ai/docs
- Download: https://lmstudio.ai/download
- Beta releases: https://lmstudio.ai/beta-releases
- Blog: https://lmstudio.ai/blog
- Claude Code integration: https://lmstudio.ai/blog/claudecode
- Source code: https://github.com/lmstudio-ai
## Related
- [[LM Studio Mobile App]]
- [[Large Language Models (LLMs)]]
- [[Claude Code]]
- [[Ollama]]
- [[HuggingFace]]
- [[Model Context Protocol (MCP)]]