Now supporting DeepSeek-V3 & Qwen2.5
Your Private AI Agents,
Always On-Device.
Run massive open-source models offline. Create custom "Pals" for coding, writing, or analysis. No data leaves your phone.
Available now on iOS. Works on iPhone and iPad with iOS 16.0 or later.
100% Private
Offline Inference
Unlimited Pals
Coding Pal
Online (Local)
Explain how transformers work in simple terms.
Imagine a transformer model like a sentence translator that pays attention to every word at once, rather than reading left-to-right...
def __init__(self):
self.attention = ...
Ask anything...
Speed
45 tokens/s
Model Size
72B Loaded