Model drop · No app update needed
A 27-billion-parameter mind, bonsai-sized.
Bonsai 27B takes a Qwen3.6-class 27B model and prunes it to a 1-bit GGUF — 27 billion parameters in 3.8 GB, running entirely on your Mac. It just appeared in your Model Store.
- 27B params · 3.8 GB on disk
- Qwen3.6-27B lineage
- 262K-token context
- Fully offline, on-device
- Apache-2.0, yours to keep
- One tap to download
Aether · Open Model Store → look for the glow
🪴
NEW
Just landed
Fits in
3.8GB
a full 27,000,000,000-parameter model, on a laptop
≈ 14× smaller · same 27B reasoning