Ternary Bonsai 2 (27B) Released on Hugging Face, Runs In-Browser with WebGPU (<6GB)

Core summary

One sentence to understand this update

Ternary Bonsai 2 (27B), a compact model derived from Qwen3.8-27B using ternary weights to achieve a size under 6GB, has been released on Hugging Face and can run locally in-browser via WebGPU.

Impact & opportunity

What this could mean

Developers can leverage this highly compressed model for efficient local and in-browser AI applications, opening new possibilities for on-device LLMs without heavy resource requirements.