PRIVATE, BY DESIGN
Your thoughts.
Still yours.
Your conversations run locally after download. The ideas you share stay on your device.
Your question→Your model→Your answer
Entirely on your phone
Read our privacy promise YOUR AI. RIGHT HERE.Runs in
Runs in
airplane mode.
Your chats stay
on your phone.
The app itself is a small download — no models are bundled inside it. Pick a model after you install, download it once, then chat with Wi-Fi and mobile data switched off. Replies are generated on your phone, and your conversations stay on your device unless you choose to share a response report. The app connects only to download models, refresh the catalog, and for account, purchase or report features you choose to use.
Available now on Google Play. Coming soon to the App Store.
Android: version 8 or later, 64-bit, and at least 3 GB of RAM for the small models or 6 GB for the 4B-class models.
Chats stay on your device On-device AI Free — no subscription Android from 3 GB RAM
Qwen 3.5
Help me find a little clarity.
Let’s make a little room.
You don’t have to figure everything out at once. Start here:
- What’s on your mind?
- What’s in your control?
- What’s one small next step?
We can take it from there.
Ask anything…
On this device. Only for you.
Search models…
Qwen 3.50.8B · 508 MB
FreeA little help, already on your phone.
Ready offline
LFM 2.52.6B · 1.6 GB
PlusReady for your next conversation.
Fits your device
Gemma 4E2B · 3.1 GB
PlusReady for your next conversation.
Fits your device
Make it yours.
A few little things, just the way you like them.
Y
Your local workspacePrivate by design
Conversations
Save chat history
Response styleBalanced
Max output tokens512
Downloads
Wi-Fi only
Your modelsOn this device
Offline AI Chat
Offline PlusChats Just between you and your phone.
App preview · illustrative conversation
Great models. A more personal home.
QwenGoogleLiquid AIMistralHugging FaceIBM
LESS NOISE. MORE SPACE.Powerful thinking.
Powerful thinking.
A beautifully simple place.
All the things you want in an AI companion.
A little less of everything else.
Your thoughts.
DOWNLOAD. THEN DISCONNECT.
Take a little
intelligence with you.
On a flight. On a train. Off the grid. Your downloaded models are ready when you are.
Your model is ready.Downloaded · Available offline
Downloads continue while you explore the app.
ROOM FOR EVERY KIND OF IDEA
From a blank page
to a little momentum.
Work through a thought, find the right words, or get a fresh perspective on a piece of code.
A first draft A simpler explanation A fresh perspective
YOUR PHONE. YOUR CHOICE.
A model for
your way of thinking.
Choose from a growing catalog. See what fits your phone, download a model, and make it your own.
37models, and growing
GOOD COMPANY, ON YOUR DEVICE.Meet your next
Meet your next
thinking companion.
Small and speedy. Thoughtful and capable.
Find a model that feels right for you.
Free
Qwen 3.5
0.8B · Qwen
0.51 GB download
Free
LFM 2.5 Instruct
1.2B · Liquid AI
0.70 GB download
Offline Plus
Gemma 4 QAT
E2B · Google
3.35 GB download
Offline Plus
Ministral 3 Instruct
3B · Mistral AI
2.15 GB download
37 models · 17 Free · 20 Offline Plus
Model availability depends on your device. Larger and experimental models need more testing and memory. On Android you need version 8 or later, a 64-bit processor, and at least 3 GB of RAM for the small models, 6 GB for the 4B-class models, and 13 GB for the largest.
Browse all 37 models in the catalogA FEW THINGS YOU MIGHT BE WONDERING.
Good questions.
Is the app free? Do I need a subscription?
The app is free to download and free to use, with no subscription required. 17 free models are available to download in the app: chat as much as you like, with no account and no time limit. Offline Plus is a separate, optional subscription for selected larger and experimental models and document chat. Free models stay free.
How big is the app itself?
Small — the app installs quickly because no AI models are bundled inside it. Models download separately, only after the first launch and only the ones you choose, over Wi-Fi if you prefer. The smallest model is about 100 MB, and the catalog shows each model’s exact download size before you start, so you decide how much space your phone gives to AI.
Does it really work without internet?
Yes, including in airplane mode. Once you’ve downloaded a compatible model, replies are generated on your phone and no connection is used. You’ll need a connection to download models, refresh the catalog, and for any account, purchase or report features you choose to use.
Where do my conversations go?
They stay in the app on your device. Our server has no chat endpoint. Conversations stay on your device unless you choose to share a response report from inside the app. You choose whether to save history, and can export or delete it in Settings.
Which models can my phone run?
On Android you need version 8 or later and a 64-bit processor. The app reads your phone’s actual memory and marks any model it can’t run. Most small models (about 1B parameters and under) need at least 3 GB of RAM, with 4 GB recommended; the two tiniest, SmolLM2 135M and Qwen 3 0.6B, run from 1.5 GB and 2.5 GB. The 2B to 3B models need 4.5 GB (6 GB recommended), the 4B-class models 6 GB (8 GB recommended), the smaller LFM 2.5 MoE and OLMoE variants 6 GB (8 GB recommended), Gemma 4 E4B and LFM 2.5 MoE Q4_K_M 10 GB (12 GB recommended), and Qwen 3.5 9B 13 GB (16 GB recommended). These minimums are conservative estimates; speed varies with your phone, model and conversation.
What’s included in Offline Plus?
The catalog currently has 17 Free and 20 Plus models. Plus includes selected larger and reasoning models. Downloaded models remain usable offline after a subscription ends. Model licenses still apply.
Where can I download the app?
The Android app is available now on Google Play. Use a Google Play button on this page to download it. The iOS app is currently in private testing; the App Store release is coming soon.
Where can I get help?
Use the Contact & support page to send a question or report a problem. Include your device, app version and model name when useful. Please don’t include passwords or private conversations.
Your next idea
can start anywhere.
Keep a little space for it.