PrismML Releases Ternary Bonsai 2 27B At A 5.9GB Footprint
PrismML is compressing a 27B model into a 5.9GB file that holds 98.2% of the base model's benchmark scores, which puts local execution within reach on graphics cards with only 8GB of VRAM.
Reporting from 1 source: GIGAZINE.
PrismML announced Ternary Bonsai 2 27B, a compressed model built on Alibaba's Qwen3.8 27B. It ships at 5.9GB, one-ninth the size of the base model, and retains 98.2% of Qwen3.8 27B's benchmark scores. PrismML lists gains in reasoning, coding, vision, and long-horizon agent performance over the earlier Bonsai 27B. The model is released under the Apache 2.0 license and published on Hugging Face. PrismML says it runs on Mac, iPhone, and iPad.
The 5.9GB figure is what separates this release from the earlier Bonsai 27B. PrismML upgraded the base from Qwen3.6-27B to Qwen3.8 27B and reports the compressed result holds 98.2% of the full model's benchmark scores, while the earlier release was marketed on running on an iPhone with heavy memory savings.
The company lists reasoning, coding, vision, and long-horizon agent performance as improved over the previous generation. Weights are on Hugging Face under Apache 2.0, and PrismML names Mac, iPhone, and iPad as target hardware.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.