🛠️ Boole-1.7B, a 1-bit LLM that outperforms Bonsai-1.7B
Summary
Boole-1.7B is a 1-bit version of Qwen3-1.7B that retains 85.9% of BF16 performance while being 13.9x smaller at 247MB. It is intended for low-latency tasks using English and non-thinking mode.
Why it’s interesting
It outperforms similar-sized 1-bit models like Bonsai-1.7B and nearly matches 1.58-bit models.
Source metrics: Points 2 · Comments 1
HN discussion · Project
Source: #HackerNews / Show HN
Summary
Boole-1.7B is a 1-bit version of Qwen3-1.7B that retains 85.9% of BF16 performance while being 13.9x smaller at 247MB. It is intended for low-latency tasks using English and non-thinking mode.
Why it’s interesting
It outperforms similar-sized 1-bit models like Bonsai-1.7B and nearly matches 1.58-bit models.
Source metrics: Points 2 · Comments 1
HN discussion · Project
Source: #HackerNews / Show HN