LU Labs

LU Labs

Your AI studio, on your PC or in the browser.

Chat, code, images and video. Free on your own hardware. No GPU? Run the same studio on ours, from €15.

Try LU Cloud in this browser

Desktop app for Windows, Mac and Linux:

WINDOWS · MACOS · LINUX · 170,000+ DOWNLOADS

Why LU Cloud as well

Some models are bigger than any desktop card.

A good desktop card runs open models up to around 30B. LU Cloud runs the ones it cannot hold, like Kimi K3 and Qwen 3.8 Max, plus video on data-center GPUs, from the same app and the same account. Local stays free.

Bigger models
47 chat models, including ones far larger than any desktop card can hold.
Faster video
5 video and 7 image models on data-center GPUs, no queue on your own card.
Works on your phone and Mac
Open the same studio in any browser. No install, no GPU.
One app · four studios

Four studios. One window.
Your machine or ours.

Run every studio on your own hardware for free. When a model is too big for your card, switch the same job to LU Cloud.

Chat view drawing an image from a request, right in the conversation

Everything you’d do in ChatGPT: ask anything, write, brainstorm, work with your own files and photos. Ask for a picture and it draws one right in the chat. Except it all stays on your computer: nothing sent to a company, nothing logged. The app even picks the right free model for your machine, so you don’t have to.

Create view making an image, with a gallery of past generations beside it

The Create tab keeps it simple: pick a look, pick a size, hit go. Make images, turn them into video. Every piece you generate lands in your own gallery. It only offers the tools your computer can actually handle, and installs each one with a single click.

Coding Agent reviewing a file and pointing out a real bug

Flip to the Coding Agent and point it at your project. Ask it to review, fix, or build. Here it caught a real bug: a session token with no expiry claim was treated as valid forever, and it explained the one-line fix in plain words. It even runs on small, light models most machines can handle.

Agent Mode writing files on its own to build a website

Turn on Agent Mode and it stops just talking and starts doing. Ask for a website and it writes the files, runs each step, and shows its work as it goes, safely, in a sandbox on your own machine.

Private by design

Nothing is ever sent automatically.

ON-DEVICE
Local runs stay on your machine
In local mode every token, pixel and frame comes from your own hardware. A job only reaches our servers when you switch it to LU Cloud yourself.
NO TELEMETRY
No telemetry in the app
No account required to run local. No analytics, no phone-home, no API keys to leak.
AUTO-CLEANUP
Processes die with the app
The ComfyUI process is auto-killed on close. Your chats survive restarts, on disk, with you.
Works with the runtimes you already have
OllamaLM StudiovLLMllama.cppComfyUItext-gen-webui
Simple pricing

Free forever self-hosted. Or let us run the GPUs.

Compare all plans
Popular guides

One model, one question, one page.

Run your first local model in minutes.

Download the LU Labs desktop app, point it at your hardware, and start generating. Private, free, and yours. Or open LU Cloud in the browser.