How to Use LU Labs Cloud on a Mac

Start with the honest bit, because it saves you a search: there is no native Mac build of the LU Labs desktop app. It ships for Windows and Linux. On a Mac, the route is the browser Studio, and this guide is how to set that up properly rather than just opening a tab and hoping.

The comparison with a local runner is on the LM Studio alternative for Mac page, and what your particular Mac can and cannot handle on its own is in what a Mac can run locally in 2026. This is the procedure.

Step 1: try it before you pay

Open the Studio demo. It needs no account, and it is enough to see whether the interface suits you. What it does not carry is the full model catalog, so do not judge the picker by it.

Step 2: choose a plan or a pack

Hosted is €19 a month with 900,000 credits, and it carries 15 chat models, 10 image models and 5 video models on one account.

A credit pack needs no subscription: €5 for 165,000 credits, €10 for 350,000, €25 for 900,000. Any pack from your first purchase, and the credits never expire.

The rule of thumb for Mac users specifically: if you are here because your Mac cannot hold a large model, take the plan, because you will be using it for chat every day. If you are here because you want an image or video studio your Mac does not have, a pack may be the better shape, since render work is bursty. The plans and packs are here.

Start on Hosted when you have decided. Card details go to Stripe and the payment is secured with 3D Secure.

Step 3: set the tab up like an app

The Studio is a web app, so treat it like one. Pin the tab in Safari or Chrome so a stray cmd+W does not lose your session. If you use several browser profiles, put the Studio in the profile you keep open all day rather than the one you close.

There is nothing to download, no runtime to keep current, and no GGUF files eating your SSD. That is the whole point of this route on a Mac, where storage is expensive and not upgradeable.

Step 4: pick a chat model on purpose

The Hosted catalog holds 15 chat models. The names you will recognise include Hermes 3 70B, Hermes 3 405B, Euryale 70B, Lunaris 8B, MythoMax 13B, Qwen 3.6 35B A3B, Qwen3 Coder 480B and Kimi K3.

Two properties are worth checking before you settle on a default:

Does it have a thinking mode? Qwen 3.6 35B A3B has a Think toggle. Hermes 3 405B has none at all. A reasoning pass costs output tokens, so leaving it on for everything is a spending decision.

Does it read images? Qwen 3.6 35B A3B does. On a Mac this matters more than you would expect, because cmd+shift+4 into the chat window is the fastest debugging loop there is.

Step 5: use the studios your Mac does not have

This is where the browser route stops being a fallback and starts being an addition. A local runner on a Mac gives you chat. It does not give you:

• Image. 10 models including Flux 2 Dev, Flux Dev and Flux Schnell, plus masked editing on Flux Dev, Qwen Image Edit, background removal, an object eraser and upscaling at 2k, 4k and 8k.
• Video. 5 models including Wan 2.2 720p, Wan 2.2 Fast, LTX-2 with audio, LTX 2.3 and HunyuanVideo 1.5, in clips of 5 or 8 s.
• LoRA training. Hosted includes 2 trainings a month. A pack on its own includes none.

Watch the credits here rather than in chat. Flux Schnell is 300 credits an image, Flux 2 Dev is 1,200, and a 5 s LTX-2 clip is 8,000. One credit is $0.00001. Chat is cheap by comparison: as a reference, Hermes 3 70B costs 0.07 credits per output token, so a €5 pack is roughly 2.3 million output tokens on it.

Step 6: get your editor on the same account

The part people miss. Go to Settings, then Cloud API keys, and create a key. It starts with lu_ and is shown once, so put it in Keychain or your password manager before you close the dialog. The key spends your plan credits and cannot read or change your account.

Base URL is https://lu-labs.ai/api/inference/v1, in the OpenAI chat completions shape, so any Mac tool that speaks that format works. Aider, LibreChat, your own scripts, or a plain curl call:

curl https://lu-labs.ai/api/inference/v1/chat/completions \
  -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"NousResearch/Hermes-3-Llama-3.1-405B",
       "messages":[{"role":"user","content":"Explain what this zsh function does, line by line."}]}'

Model ids are the catalog ids, for example NousResearch/Hermes-3-Llama-3.1-405B and Qwen/Qwen3.6-35B-A3B.

Step 7: decide what stays local

Do not throw away your local setup. If you already run LM Studio or Ollama on your Mac, keep them for the offline everyday: on a plane, on bad hotel Wi-Fi, or for anything that should never leave the disk. Our comparison of Ollama, LM Studio and LU Labs covers where each one fits, and the Mac local AI guide covers the local install itself.

Most people who use both settle into the same split: local for the private and the offline, cloud for the models that never fit and for the image and video work a laptop cannot do.

Troubleshooting

The picker looks short. You are in the demo Studio. Sign in to the paid account.

Safari asks about downloads when I save a render. Normal browser behaviour. Allow downloads from the site once and it stops asking.

The reply pauses before streaming. Shared hardware, so a request can queue a few seconds at busy hours. It clears on its own.

My API key returns 401. Keys are shown once and cannot be recovered, only revoked and replaced. Send it as Authorization: Bearer lu_....

Where is the Mac app. There is not one. Windows and Linux only today, and the browser Studio is the Mac answer rather than a temporary workaround.

A render was refused. Hosted image and video follow the Terms of Service. Content involving minors is refused everywhere, always.

Account data is hosted in the EU and you can delete it yourself. We never train on your data and we never sell it. The work runs on managed NVIDIA H100, A100 and B200 class GPUs, which is why an Intel MacBook and an M-series one perform identically here: your Mac is the screen, not the engine.