LM Studio
A desktop app to find, download and chat with open AI models offline
- Made by
- LM Studio
- Category
- Model
- Last change
- Oct 29, 2026
Changelog
October 2026
- New feature
LM Studio now supports the oMLX engine, which users can download and use with MLX models they already have.
What this means for you: LM Studio can now use the oMLX engine. You download the engine inside the app and use it with MLX models you already have, which are models in the format made for Apple computers.
How to use it: In LM Studio, download the new oMLX engine, then use it with your existing MLX models.
I use LM Studio on my Mac and have some MLX models downloaded. I just learned it now supports the oMLX engine. Explain in plain words what an engine does in LM Studio, and give me a simple checklist for trying oMLX with my existing models. My Mac is: [model and memory].
- New model
LM Studio added Liquid AI's LFM3 model, which fits in about 1 GB of memory.
What this means for you: LM Studio, a desktop app for downloading and chatting with AI models offline, now has Liquid AI's LFM3 model. It fits in about 1 GB of memory, so it can run on a computer without much spare memory.
How to use it: Open the LFM3 model page at lmstudio.ai/models/lfm3 to download it into LM Studio.
LM Studio just added Liquid AI's LFM3 model, which fits in about 1 GB of memory. Help me decide whether it suits what I do, and what to test first. My computer has [amount] of memory and I mostly want to use it for [tasks, e.g. summarizing notes or drafting emails].
- New model
LM Studio added Flint from Recursive, a 40B coding model with a 262K context window that needs about 24GB to run locally.
What this means for you: Flint is a coding model you can download and chat with inside LM Studio, with everything running on your own computer. It is a 40B model (40 billion parameters, the internal settings that make up a model), and it can hold about 262K tokens (small chunks of text) in mind at once, so you can give it a large amount of code.
Why it matters: It comes six days after LM Studio added NVIDIA's Nemotron 3.5 Super VL on Oct 15, which needs about 70GB of RAM. Flint needs about 24GB, so it fits on a smaller machine.
How to use it: Open LM Studio and search for Flint from Recursive. It needs about 24GB to run locally. It is licensed Apache 2.0 and was added on day 0, the day it was released.
I just learned LM Studio added Flint, a 40B coding model from Recursive that runs offline and needs about 24GB to run locally. Help me decide if my computer can run it and whether it suits my work. My computer: [model, RAM]. What I want help coding: [describe tasks].
- New model
LM Studio added NVIDIA's Nemotron 3.5 Super VL, a 120B model with vision input that needs about 70GB RAM locally.
What this means for you: You can run NVIDIA's Nemotron 3.5 Super VL on your own computer through LM Studio, with no cloud service involved. It accepts images as input and is built for reasoning, visual understanding and tool use.
How to use it: Find it in LM Studio at lmstudio.ai/models/nemotron-3.5-super-vl. It needs about 70GB of RAM to run locally.
I'm thinking about running NVIDIA's Nemotron 3.5 Super VL in LM Studio. It needs about 70GB of RAM and accepts images. Help me work out whether my computer can run it and what I could use it for. My computer has: [chip, RAM, storage].
September 2026
- New feature
LM Studio added the Inco Splash inference engine as an experimental backend, with Qwen3.8-27B running at up to 144 tokens per second on Apple Silicon.
Related tools
- DeepSeekChinese lab's open-weight models and free chat app
- OllamaDownload and run open AI models like Llama, Gemma and Qwen on your own computer
- OpenRouterOne account and one API to use hundreds of AI models from every major lab
- ReplicateRun open AI models in the cloud with one click or one line of code
- HunyuanTencent's open AI models for video, images and 3D
- KimiMoonshot's open-weight models and chat app
