mu
AI
New chat
© Kancamu Project
New chat
Export .md
Choose model
AI that runs in your browser
Models run directly on your device's GPU through WebGPU. No server, and your chats stay in this browser.
Model
Load model
Change model
Preparing…
Attached image
Remove
Send
Stop
Enter to send · Shift+Enter for a new line · AI can be wrong, double-check important info.
Choose AI model
All models run in your browser through WebGPU and are downloaded from Hugging Face once. Sizes are the approximate first download. Larger models are smarter but need more GPU and memory.
Close
Settings
System prompt
Base behavior of the AI in every chat. Leave empty for none.
Max response tokens
Length limit of a single reply (64–4096).
Creativity (temperature):
0 = always the same and consistent, higher = more varied.
Messages sent as context
Only the last N messages are read by the model. Smaller = faster and uses less memory (2–60).
Theme
Dark
Light
WebGPU
Checking…
Data
Chats are stored only in this browser. Export a backup, restore one, or wipe everything.
Export all
Import
Delete all
Reset
Cancel
Save