a billion parameters in your pocket
chat with private and local large language models
the simplest way to use private LLMs
chat with private and local large language models
works fully offline when you don't have internet
runs models on-device optimized for Apple silicon
works on every platform
personalize
adjust the theme, fonts, and system prompt.
Shortcut
chat or get an output from a local model to use with other actions.
TestFlight
install the TestFlight to get the latest features and models.
install
free, open source, private
free
open source
private
tech specs
fullmoon
chip
Apple silicon
graphics
Metal 3
array framework
Swift MLX
platforms
iOS iPadOS macOS visionOS
models
Llama-3.2-1B-Instruct-4bit Llama-3.2-3B-Instruct-4bit DeepSeek-R1-Distill-Qwen-1.5B-4bit DeepSeek-R1-Distill-Qwen-1.5B-8bit
source
Llama-3.2-1B-Instruct-4bit
params
193M
tensor type
FP16 • U32
precision
4-bit
base model
size
0.7 GB
Llama-3.2-3B-Instruct-4bit
params
502M
tensor type
FP16 • U32
precision
4-bit
base model
size
1.8 GB
DeepSeek-R1-Distill-Qwen-1.5B-4bit
params
278M
tensor type
FP16 • U32
precision
4-bit
base model
size
1.0 GB
DeepSeek-R1-Distill-Qwen-1.5B-8bit
params
500M
tensor type
FP16 • U32
precision
8-bit
base model
size
1.9 GB