LM Studio

MCP Catalog and Toolkit Learn about Docker's MCP catalog on Docker Hub

🚀 LM Studio now works out of the box with the Docker MCP Toolkit!

Skip the messy configs—connect MCP servers in one click to LM Studio.

🛠️ Build agents easily & securely with Docker.
🔗 docs.docker.com/ai/mcp-catal...

#DockerAI #MCP #DevTools #LMStudio

lms chat by mayfer · Pull Request #227 · lmstudio-ai/lms LMS cli chat mode Examples: lms chat Start chat REPL with default loaded model lms chat modelname Start chat REPL with specific model lms chat -p "Your prompt" Print lms response ...

yuppp :) github.com/lmstudio-ai/...

👀

Apple Reveals New Mac Studio Powered by M4 Max and M3 Ultra Today, Apple revealed the new Mac Studio featuring both M3 Ultra and M4 Max options. It’s an odd assortment on its face, so let’s take a closer look at what’s going on. As with the original Mac Studio...

M3 Ultra Mac Studio:

"Up to 16.9x faster token generation using an LLM with hundreds of billions of parameters in LM Studio when compared to Mac Studio with M1 Ultra"

😳

www.macstories.net/news/apple-r...

💯! @mcuban.bsky.social you should try it out :-)

LM Studio with DeepSeek R1 Distill Llama 8B Q4_K_M correctly answering a question about how many O's there are in Potato

LM Studio Mission Control, showing a searchable list of Models for download.

I was expecting this to take me a couple of hours to set up (with me getting annoyed about undocumented requirements on Github), but no, a few mins including the mini-model download.

LM Studio is very beginner friendly.

🥔⚖️⭕⭕

E tai, o Modelo de IA Deepseek rodando no meu MacBook Air M3 16Gb a 6 tokens por segundo LOCALMENTE.

É muito simples, basta baixar o LM Studio , procurar pelo modelo, baixar (5GB) e rodar.

Vc tem uma IA de ponta 100% offline rodando no seu computador

Surreal.

🚀 Zed v0.170 is out!

Zed's assistant can now be configured to run models from @lmstudio.

1. Install LM Studio
2. Download models in the app
3. Run the server via `lms server start`
4. Configure LM Studio in Zed's assistant configuration panel
5. Pick your model in Zed's assistant

Run DeepSeek R1 locally with @lmstudio-ai.bsky.social 🥰🥰🥰

Guess who just figured out how to interact with a local #LLM model in #rstats?

👉This guy!👈

(I did this via @lmstudio-ai.bsky.social using @hadley.nz's 'ellmer' package. Happy to share how I did it if people are interested).

Really enjoying local LLMs.

LM Studio appears to be the best option right now. Its support for MLX-based models means I can run LLama 3.1-8b with a full 128k context window on my M3 Max MacBook Pro with 36 GB.

Great for document chat, slack synopsis, and more.

What is everyone else doing?

🔥

Screen capture of VSCode running Cline with LLM generated Go code.

VSCode with #Cline plugin connected to a local LM Studio running Qwen 2.5 14B LLM on an M4 pro...

Programming prompt compiled and ran the second time after it self-corrected, all tests pass. Code generation took less than a minute to complete.

😳 @lmstudio-ai.bsky.social @vscode.dev

Issues · lmstudio-ai/lmstudio-bug-tracker Bug tracking for the LM Studio desktop application - Issues · lmstudio-ai/lmstudio-bug-tracker

Thanks. A few more datapoints will help.

1. Which quantization of model is it for either case?
2. When you load the model in LM Studio, what is the % of GPU offload you define?

It might be easier to go back and forth in a github issue: github.com/lmstudio-ai/...

Thanks!

Should be as fast or faster. Pls share more details and we’ll help debug

a computer screen with an open program, likely written in JavaScript rather than Python. The program appears to be part of the "Codestral Model" and is being used for solving "adventofcode" challenges. The screen has a blue background and black text, making it easy to read.

Codestral on LM Studio lowkey slays

Llama 3.2 1B llama • Meta • 1B

Try a very small model like llama 3.2 1B lmstudio.ai/model/llama-...

Very cool! 🦾 Did you try using local LLMs call as a part of the numerical / analytical process itself, or for R code gen?

@martinctc.bsky.social with a cool blog post showing how to combine local LLM calls with your R code

@lmstudio-ai.bsky.social detected my vulkan llama.cpp runtime out of the box, absolutely magical piece of software ✨

Didn't think I have a chance with a smol 12GB 4070 to run any interesting LLM locally.

But Qwen2.5-14B-Instruct-IQ4_XS *slaps*. It's no Claude, but I'm amazed how good it is.

(Also shout out to @lmstudio-ai.bsky.social - what a super smooth experience.)

We’re hoping for as much feedback as possible from devs who have been using the OpenAI Tool Use API 🙏

📣🔧Tool Use (beta)

Are you using OpenAI for Tool Use?

Want to do the same with Qwen, Llama, or Mistral locally?

Try out the new Tool Use beta!

Sign up to get the builds here: forms.gle/FBgAH43GRaR2...

Docs: lmstudio.ai/docs/advance... (requires the beta build to work)

I'm really happy with Tulu 3 8B. Is nice in the hosted demo (playground allenai org), but also quantized running locally on my mac in LM studio D. Feels like a keeper.

LM Studio is magic. You’re going to want a very beefy machine though.

Never realised how easy it is to run LLMs locally. Thought I'd spend at least half a day to get it up and running, but with LM Studio it took me less than 15 min today. I feel like a grampa saying "I used to spend days to get Caffe running, you don't know how easy it is today!" #databs

lm studio screenshot, i ask the assistant "Who is haliey welch" and the assistant responds in a loop "HAWK TUAH GIRL! HAWK TUAH GIRL! HAWK TUAH GIRL!"

okay i didn't believe the hype but it's kinda crazy just how good LLMs have gotten over the past few years...

Wow at #MsIgnite looking at the FIRST build of a local AI model running entirely in a snapdragon NPU #AI TikTok video by Scott Hanselman

WOW check this out it’s LMStudio.ai @lmstudio-ai.bsky.social running ENTIRELY locally on an NPU from Qualcomm www.tiktok.com/t/ZTYYaN6cn/