Show HN: Boole-1.7B, a 1-bit LLM that outperforms Bonsai-1.7B
precisionnumerics.aiIntroducing Boole-1.7B, a 1-bit version of Qwen3-1.7B.
Boole-1.7B retains 85.9% of BF16 performance using 247MB (13.9x smaller). It outperforms 1-bit Bonsai-1.7B and similar-sized LFM models, and nearly matches 1.58-bit Bonsai-1.7B.
It’s intended for low-latency tasks using English and non-thinking mode. To learn more,
- Blog: https://precisionnumerics.ai/blog/boole-1p7b
- Weights: https://huggingface.co/precision-numerics/Boole-1.7B
- Chat in browser: https://precisionnumerics.ai/chat/
We're going to ship more capable models soon for the open-source community. Let us know what models/capabilities you want to see next!
Honestly, a local model that can run on CPU with reasonable speeds and have a good tool call success rate would be great to have. Think about it, I carry my laptop everywhere with me but I don't enable my Nvidia graphics card because I want to be able to use it for more than 30 minutes. I would love to have a model that can quickly put calendar entries, create todo lists and expand upon my Obsidian notes via Web Search, but unfortunately I have to rely on third party providers to achieve this for the time being. I don't expect local 1 bit models to ever code or do complex tasks but basic tool calls and language functionality would be awesome to have fully offline everywhere.