AI BriefingPrecision Numerics Releases Boole-1.7B, a 1-bit Qwen3 Model Outperforming Bonsai-1.7B
Key point
The new model achieves 85.9% of the BF16 baseline score while being 13.9× smaller in file size.
Details
Precision Numerics has launched Boole-1.7B, a 1-bit quantized version of the Qwen3-1.7B model. The release focuses on extreme compression for local deployment, with packed weights totaling 247 MB, making it 13.9× smaller than the original BF16 model.
Performance and Benchmarks
In a 10-task evaluation, Boole-1.7B achieved 85.9% of the average score of the full BF16 model. The release claims this performance outperforms 1-bit Bonsai-1.7B and nearly matches the 1.58-bit Bonsai-1.7B.
Deployment and Availability
The model is available for local execution in a web browser (optimized for English) and supports integration with vLLM and llama.cpp. Weights are hosted on Hugging Face, and demo code is available on GitHub.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Samsung Labs Releases LittleBit-2 for Sub-1-Bit LLM CompressionAgent Standup: Open-source tool visualizes Claude Code sub-agents as a live team meeting
Talorys: Open-Source Personal AI Agent for Cloudflare Free Tier