Announcement
Bonsai 2 adds 2B, 4B and 9B models for everyday devices
PrismML says the Qwen3.5-based ternary models range from 0.43GB to 2.0GB and are available under Apache 2.0.
TLDR
PrismML says it released three Qwen3.5-based Bonsai 2 ternary models on October 15, 2026: the 2B is 0.43GB, the 4B is 0.95GB and the 9B is 2.0GB. The company designed them for local assistants on personal computers, phones and wearables. It says the 9B uses roughly 40% fewer tokens on standard instruction-following and everyday math tasks.
Combined views
—
2 Sources, first seen ago
