OpenBonsai开源三值量化转换框架
OpenBonsai 开发者发布 Apache-2.0 框架，可将 BF16 模型及微调模型转换为兼容 Prism PTQ1_0 格式的三值 GGUF。开发者称，Qwen3.5-9B 从约 18 GB 压缩至 1.997 GB，同一套流程还生成了 1.8 GB 的 Qwen3-8B 文件。初步结果显示，其模型在多数已报告的推理与编程测试中低于 Prism Ternary Bonsai 2，项目目前仍被定位为研究预览。
OpenBonsai 公开了此前不可用的 BF16 到三值 GGUF 转换流程及校准、量化、评测脚本和失败实验，但结果由开发者自行报告，且与 Prism 模型仍存在质量差距。
I've been experimenting with tiny models for WebBrain, my open-source browser agent. The long-term goal is to run as much intelligence as possible locally, ideally through WebGPU.
One approach I've been particularly interested in is Prism ML's Ternary Bonsai: compressing a 9B model to roughly 2 GB using ternary weights (-1, 0, +1).
Prism released the weights and inference runtime, but not the complete BF16-to-ternary compression pipeline. You can fine-tune their existing ternary models, but there's no publicly available way to compress your own BF16 fine-tune into their format.
So I tried building one.
Introducing OpenBonsai, an Apache-2.0 framework for post-training ternary quantization, compatible with Prism's PTQ1_0 GGUF format.
How it works
A recent Intel Labs China paper, "Attend to Your Own Thoughts" (Wang et al.), inspired part of the approach: using the model's own reasoning traces for calibration rather than relying exclusively on generic text.
#OpenBonsai#Prism ML#Ternary Bonsai 2#Qwen3.5-9B#Qwen3-8B#PTQ1_0 GGUF#llama.cpp#Intel Labs China