LM Studio — A free desktop app for discovering, downloading, and running open-weight large language models locally via llama.cpp and MLX, with optional pay-as-you-go cloud inference.
By Noahlast updated pricing verified
LM Studio screenshots & video
LM Studio in 20 seconds
Best for: developers and privacy-conscious users who want to run open-weight LLMs locally with an OpenAI-compatible local API
Pricing: Free for local model use; optional pay-as-you-go cloud inference billed per million tokens (from about $0.13 to $3.00+ per million input tokens depending on model).
- LM Studio is free for local model use, with no account required, on Windows, macOS, and Linux.
- LM Studio runs local inference via the llama.cpp and MLX engines.
- LM Studio's optional pay-as-you-go cloud tier bills per million tokens, from roughly $0.13/million input tokens for smaller models to $3.00/million for larger ones.
- LM Studio includes a built-in OpenAI-compatible local API server for integrating with other tools.
- LM Studio holds a 4.7/5 rating from 4 reviews on Product Hunt.
Limitations: Larger local models require substantial RAM or a capable GPU.; Bionic Pass subscription pricing not yet published.; The app itself is closed-source, unlike the open-source engines it runs on.
Top alternatives: Ollama, Open WebUI, Jan, GPT4All, Msty — full list
What is LM Studio?
LM Studio is a free desktop application that lets developers and hobbyists discover, download, and run open-weight large language models entirely on their own computer, using the llama.cpp and MLX inference engines rather than calling a cloud API. Because inference runs locally, LM Studio is positioned around privacy and offline use: chats, documents, and voice transcription can all be processed without data leaving the device. Beyond basic local chat, LM Studio includes Bionic Agent (an agentic mode for coding and document tasks with automatic saving), a built-in local server that exposes an OpenAI-compatible API for other apps to call, and LM Link for syncing across up to 5 devices. LM Studio itself is free, closed-source freeware — distinct from the open-source llama.cpp and MLX runtimes it's built on — with an optional pay-as-you-go add-on for cloud-hosted models when local hardware isn't sufficient, billed per million tokens with zero data retention.
LM Studio is a free desktop application for discovering, downloading, and running open-weight large language models locally on Windows, macOS, and Linux, powered by the llama.cpp and MLX inference runtimes, so models run entirely on-device without sending data to a cloud API. The core app — including local model chat, an agent mode called Bionic Agent, local voice transcription, and a developer API server compatible with OpenAI's format — is free with no account required, while an optional pay-as-you-go tier bills per-million-tokens for hosted cloud models, priced from roughly $0.13 per million input tokens for smaller models up to $3.00 per million for larger ones, billed with zero data retention. LM Studio holds a 4.7/5 rating from 4 reviews on Product Hunt, and its biggest practical limitation is hardware: running larger local models well requires substantial RAM or a capable GPU, so users on modest hardware are effectively pushed toward smaller quantized models or LM Studio's optional paid cloud inference.
Standout feature: Runs open-weight models entirely on-device via llama.cpp and MLX with zero cloud dependency, while still exposing an OpenAI-compatible local API server so existing tools built for cloud LLM APIs can point at a local model instead.
Biggest weakness: Running larger local models at usable speed requires substantial RAM or a capable GPU, so users on modest hardware are effectively limited to smaller quantized models or pushed toward the optional paid cloud-inference tier; the Bionic Pass subscription tier also has no published pricing yet.
Pros
- Entirely free for local model use with no account or subscription required.
- Runs models on-device via llama.cpp and MLX, keeping data off the cloud by default.
- Built-in OpenAI-compatible API server lets other tools call local models as a drop-in cloud API replacement.
- Optional pay-as-you-go cloud tier fills the gap when local hardware can't run larger models, without forcing a subscription.
Cons
- Running larger models well requires substantial RAM or a capable GPU; modest hardware is limited to smaller quantized models.
- Bionic Pass subscription tier has no published pricing yet ("coming soon").
- LM Studio itself is closed-source freeware, not open-source, despite being built on open-source inference engines.
- Very thin public review base (4 total on Product Hunt) makes it hard to gauge broad user sentiment.
LM Studio pricing
Free · verified from the vendor's pricing page
| Plan | Price | Notes |
|---|---|---|
| Free (local) | Free | Bionic Agent, local LLM inference via llama.cpp/MLX, local voice transcription, web search, LM Link for up to 5 devices; no account required. |
| Pay-as-you-go (cloud inference) | Custom | Billed per million tokens for hosted models (roughly $0.13/million input tokens for smaller models up to $3.00/million for larger ones); zero data retention, US-based inference. |
| Bionic Pass | Custom | Pricing and plan details not yet published ("coming soon" per lmstudio.ai). |
What reviewers say
Scores belong to their platforms; we display them with attribution and never fold them into our own ratings.