iPhone and iPadiOS 18.2 or laterComing to the App Store
A private AI chat that runs on your iPhone.
Download an open model once and chat with it offline. Ask questions, summarize a PDF, read a menu in another language. Your messages stay on the device.
- open models to choose from
- 20open models to choose from
- smallest download
- 190 MBsmallest download
- accounts, trackers or ads
- 0accounts, trackers or ads
Answered in airplane mode.
What it does
Ask about photos, files and drafts without sending them anywhere.
How it works
Download a model once, then use it offline.
Step 1
Pick a model
If your iPhone has Apple Intelligence and iOS 26, you can start with Apple’s built-in model, no download needed. Or choose one from the catalog; the app marks the ones recommended for your device.
Step 2
Download it once
The app shows the size and the free space it needs before you start. If the download is interrupted, it resumes. Updates only download the files that changed.
Step 3
Chat offline
The model runs on your iPhone’s GPU with Apple MLX. You don’t need an account, and after the download you don’t need a connection either.
Models
Which models make sense on your iPhone?
It comes down to memory. The app checks how much your device has and recommends models that leave room for iOS and your other apps. Pick your iPhone to see what it would suggest.
iPhone 16 has 8 GB of memory. 15 of 20 models run on it; 5 need more memory.
Best overall
Granite 4.2 3B
Writing, questions, summaries
2.1 GB download
Photos & images
Qwen 3.5 4B
Reads photos and screenshots
3.1 GB download
Fast & light
LFM 2.5 1.2B Instruct
Quick answers, smaller download
660 MB download
Everything that runs on iPhone 16, by download size
- Gemma 3 270M InstructText190 MB
- Qwen 3.5 0.8BText + images650 MB
- LFM 2.5 1.2B InstructText660 MB
- Gemma 3 1B InstructText770 MB
- SmolVLM2 500M Video InstructText + images1.0 GB
- MiniCPM5 2BText1.4 GB
- LFM 2.5 VL 1.6BText + images1.5 GB
- Qwen 3.5 2BText + images1.8 GB
- Qwen 3 VL 2B InstructText + images1.8 GB
- Granite 4.2 3BRecommendedText2.1 GB
- Gemma 3 Text 4B InstructText2.6 GB
- Gemma 3 4B Vision InstructText + images3.0 GB
- Qwen 3.5 4BRecommendedText + images3.1 GB
- Qwen 3 VL 4B InstructText + images3.1 GB
- Gemma 4 E2B InstructText + images3.6 GB
What the Models screen shows on iPhone 16. Apple Intelligence also appears here on iOS 26.
Details
Also in the app.
Storage
Settings shows how much space your models take and how much is free. Delete a model to get the space back; you can download it again later.
Heat
Running a model works the chip hard. If the device gets too warm, the app pauses and tells you, and picks up again once it cools down.
Personalisation
Tell it your name, what you do and how you like answers. Each conversation can also have its own instructions and temperature.
Stop, continue, regenerate
Stop an answer halfway, continue one that hit the length limit, or regenerate it. You can also edit a message and send it again.
Siri and Shortcuts
Say “Ask LMLocal” or use it in Shortcuts. Your question opens in a new chat and is answered on the device.
Conversations
Search and rename your chats, or export one as Markdown. They live in a side drawer on iPhone and in a sidebar on iPad.
Privacy
Nothing you type leaves your iPhone.
The model is a file on your iPhone and runs on its chip, so your conversations never go through a server. There are no accounts and no analytics. The app only connects to Hugging Face, to download or update models.
| Your messages | Generated and stored on the device |
|---|---|
| Photos and files | Stored inside the app, on the device |
| Model files | Stored on the device until you delete them |
| Account | None |
| Analytics | None |
| Internet | Only to download or update a model |
FAQ
Common questions.
Is it as smart as ChatGPT?
No. These are small models, from 270M to 12B parameters, so they fit in a phone’s memory. They work well for summaries, rewriting, describing images and everyday questions. For expert knowledge or very long documents, a large cloud model will do better.
Which iPhone do I need?
Any iPhone or iPad with iOS 18.2 or later. Memory decides how many models it can run: 5 of 20 with 4 GB (iPhone 12, 13), 15 with 8 GB (iPhone 15 Pro, 16) and 19 with 12 GB (iPhone 17 Pro, Air). The largest one needs 16 GB.
How much storage does it use?
Each model is a single download between 190 MB and 11 GB. Apple’s built-in model needs no download at all.
Does it drain the battery?
While it’s answering, yes: the GPU works hard, so long sessions use battery and warm the phone up. If it gets too warm, the app pauses until it cools down.
Does it work without internet?
Yes. You need a connection to download or update a model. After that, chatting works in airplane mode.
Where do the models come from?
From Hugging Face, in MLX format. Each one is pinned to a specific version, so everyone downloads the same files, and the app links to its Hugging Face page.
Coming soon to the App Store.
For iPhone and iPad, iOS 18.2 or later, in English and Spanish. Send us an email and we’ll let you know when it’s out.