跳到正文
RCreddit.com·
暂不在当前实时榜单

bonsai's document reveal how much cherry picked their headlines are

AI 摘要

Bonsai's documents reveal discrepancies in their performance claims, particularly regarding the Ternary Bonsai 2 27B model. While Bonsai initially claimed 98.2% intelligence retained, their own evaluations show Ternary Bonsai 2 27B achieving 52.8 and 60.8 on Terminal-Bench 2.1 and SWE-bench Verified, respectively. This performance is compared to Qwen3.5-27B's 69.7 and 80.6, indicating that Ternary Bonsai 2 27B retains roughly three quarters of the full-precision performance on both benchmarks, rather than the higher figure initially suggested.

为什么是这条

This report uses Bonsai's own documents to compare its Ternary Bonsai 2 27B model's actual performance against its advertised claims, unlike other reports that only cite the company's public statements.

时间与来源

时间显示为 UTC

显示时区:UTC

本地时区尚不可用,暂时显示 UTC。

收录当时偏移:UTC+02026年9月18日 14:00 UTC

收录
2026年9月18日 14:00
来源类型
开发者社区

讨论趋势

→ 平稳
最近 24 小时与此前 24 小时的快照均值对比 · 7 天曲线

百分比基于采集到的讨论信号,不代表新增评论数或独立参与人数。曲线仅用于同一话题在不同时段的比较。

本站未收录正文。

前往源站阅读 →
来源·reddit.com