Original Summary

Introducing Boole-1.7B, a 1-bit version of Qwen3-1.7B.<p>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.<p>It’s intended for English, non-thinking use. To learn more,<p>- Blog: <a href="https:&#x2F;&#x2F;precisionnumerics.ai&#x2F;blog&#x2F;boole-1p7b" rel="nofollow">https:&#x2F;&#x2F;precisionnumerics.ai&#x2F;blog&#x2F;boole-1p7b</a> - Weights: <a href="https:&#x2F;&#x2F;huggingface.co&#x2F;precision-numerics&#x2F;Boole-1.7B" rel="nofollow">https:&#x2F;&#x2F;huggingface.co&#x2F;precision-numerics&#x2F;Boole-1.7B</a> - Chat in browser: <a href="https:&#x2F;&#x2F;precisionnumerics.ai&#x2F;chat&#x2F;" rel="nofollow">https:&#x2F;&#x2F;precisionnumerics.ai&#x2F;chat&#x2F;</a><p>We&#x27;re going to ship more capable models soon for the open-source community. Let us know what models&#x2F;capabilities you want to see next!

  • 情报分类：技术学习与提效
  • 分类依据：内容涉及技术、AI、软件工具或工程实践
  • 信息来源：Hacker News 新项目
  • 发布时间：2026/10/12 23:49:48