FAQ
LU Labs FAQ: privacy, quotas and billing
Questions, answered, including the difference between local and hosted.
General
What is LU Labs?
A privacy-first desktop app for local AI. It runs local chat, a coding agent, and image & video generation entirely on your own machine, plus optional hosted plans if you would rather not manage a GPU.
Is there anything between me and the model in local mode?
No. In local mode the model you load defines its own behaviour, and there is no cloud service in the loop: nothing is logged, nothing leaves your machine.
Which models can I run?
Anything your local backend supports. LU Labs auto-detects 12 runtimes including Ollama, LM Studio, vLLM and llama.cpp, and drives ComfyUI for images and video.
Does every chat model work in Agent and Code?
Yes. On hosted plans every chat model in the catalog can call tools, so Agent mode and the coding agent run on all of them. The model picker marks each row, so you can see before you pick.
What are the limits on hosted image and video?
Your monthly credits and our Terms of Service. Content involving minors is refused everywhere, always. Want no server in the loop at all? The image and video models also run locally, and that path is free.
Privacy
Does anything leave my machine?
In local mode, no: not prompts, files, images or telemetry. Nothing is ever sent automatically. Hosted plans process on managed NVIDIA H100, A100 and B200 class GPUs, but only the request you explicitly send.
Do you collect analytics?
No phone-home, no analytics SDK, no crash pings on local mode. It runs entirely on your machine, so nothing is sent anywhere.
What happens to background processes?
The ComfyUI process is auto-killed when you close the app, so nothing keeps running or listening after you quit. Your chats are stored on your own disk and survive restarts.
Pricing & billing
Which Flash models are included on my plan?
Hosted includes 11 Flash models without credits; Pro and Max include 12. Unlimited prompt counts, up to 500,000 combined input and output tokens per account per UTC day in app sessions. All 47 chat models remain available on every plan. GLM 5.3 Flash uses credits on Hosted and is included in the daily allowance on Pro and Max. API keys and accounts without an active subscription always use credits.
Are images, video and API separate allowances?
No. Images, video, API and metered text use one shared monthly credit wallet. The daily Flash allowance is separate. Exact per-model quantities and current plan details are published at https://lu-labs.ai/product-facts.
Is the app really free?
Yes. Self-hosting is free forever. You only pay if you choose a hosted plan so we run the GPUs for you.
Monthly or yearly?
Both. Yearly billing gives you two months free versus paying monthly. You can switch or cancel at any time from your account.
Hosted vs self-host
Why pay if local is free?
Convenience and hardware. Hosted plans need no GPU, no setup and no maintenance: the same studios, running on managed NVIDIA H100, A100 and B200 class GPUs, in your browser.
What GPUs do the hosted plans run on?
Managed NVIDIA data-center GPUs. Chat and code run on H100 and A100 class hardware; image, video and voice run on H100 and B200 class hardware. You never rent, set up or maintain a card yourself, and you only pay per generation through credits.
Do I get a GPU, or can I upload my own model?
No. A hosted plan is not GPU rental. You generate from our curated catalog of open-weight models running on managed GPUs; there is no raw GPU to manage and you don't upload your own base model. You can train LoRA style adapters on top of our base models. If you need a specific model we don't host, self-hosting is free and runs anything your own hardware supports.
Are hosted results different?
Same studios, and since September 2026 the same models. Every plan and every credit pack carries the whole open-weight chat catalog, plus every image and video model. What the tiers differ in is the monthly credit budget, which is how much you can spend on that catalog.
Can I move between them?
Anytime. There is no lock-in: export your work and self-host, or start hosted and drop to local whenever you like.