Installer lm-studio avec Homebrew Cask

Liens du projet

• Page d'accueil
• Page du gestionnaire de paquets

Commandes et alias

Aucun alias exécutable n'a été trouvé dans la base locale des paquets.

Historique du projet

LM Studio emerged in 2023 as a desktop application for finding, downloading, and running local language models. Its early design combined an offline-capable chat interface, Hugging Face model discovery, model management, and an OpenAI-compatible local server.

Version 0.3.0 in 2024 added document chat, internationalization, structured output support, and network serving. The project then expanded from a GUI application into a wider local-AI stack with lms, SDKs, open-source inference engines, and llmster for headless deployment.

Historique d'adoption

LM Studio gained adoption during the rapid expansion of downloadable open-weight language models by packaging model discovery, quantized-model downloads, chat, and hardware-aware inference in a desktop interface. Its OpenAI-compatible local server made it useful beyond interactive chat because existing clients could be pointed at a local endpoint.

The project subsequently added an lms command-line tool, Apple MLX support, document chat, headless operation, Python and TypeScript SDKs, tool calling, MCP support, authentication, parallel serving, and a native REST API.

Modes d'utilisation

Users search for compatible models, download quantized variants, load them into memory, and chat locally through the desktop interface. The application can also expose REST and OpenAI- or Anthropic-compatible endpoints for local applications.

The lms CLI supports model downloads, loading and unloading, chat, daemon management, and server control. Named configuration presets are JSON files stored under ~/.lmstudio/config-presets on macOS and Linux.

Local API authentication is optional. Tokens are created and managed in the Developer page and are not shown again after creation; official documentation does not expose a stable token-storage file, so credentials-file-location is null.

Pourquoi les passionnés de paquets s'y intéressent

LM Studio is a prominent desktop distribution channel for local LLM runtimes and quantized model files. It lowers the setup cost of llama.cpp- and MLX-based inference while exposing enough CLI and API surface for scripting and application development.

For package enthusiasts, it sits at the intersection of heavyweight desktop apps, GPU-specific inference runtimes, very large model artifacts, local API compatibility, and reproducible model presets. The desktop application itself is not published as a single open-source repository, although several SDKs, engines, the CLI, and documentation are developed openly.

Chronologie

• 2023: LM Studio began its public life as a desktop local-LLM application.
• 2024: The lms CLI and LM Studio 0.3 arrived, broadening automation, document chat, and local-server capabilities.
• 2025: Official Python and TypeScript SDKs reached their initial stable releases.
• 2026: LM Studio 0.4 added llmster, parallel inference, a native stateful REST API, and a refreshed interface.

Related projects

• llama.cpp provides one of LM Studio's supported local GGUF inference backends.
• LM Studio's mlx-engine supports Apple MLX models on Apple Silicon.
• lms is LM Studio's open-source companion CLI.
• lmstudio-python and lmstudio-js expose LM Studio capabilities to applications.

Sources

• https://github.com/lmstudio-ai
• https://lmstudio.ai/blog/0.4.0
• https://lmstudio.ai/blog/introducing-lmstudio-sdk
• https://lmstudio.ai/blog/lms
• https://lmstudio.ai/blog/lmstudio-v0.3.0
• https://lmstudio.ai/docs/app