New todaydeepseek-v4.1IMAGE

Artificial Analysis v5 with private tests widens open-vs-closed gap from 3 to 12 points

Intelligence Index v5 moves 60% of tests to private sets, widening the open-vs-closed gap from 3 to 12 points; top open-weight scores are DeepSeek V4.1 Flash and GLM-5.3 at 32, still under Claude Haiku 5.5's 36.

松子 Songzi

@_SongzioeveArchived 11 hours agoOriginally posted 2026.10.28

Artificial Analysis v5 with private tests widens open-vs-closed gap from 3 to 12 points (image 1 of 2)

Artificial Analysis v5 with private tests widens open-vs-closed gap from 3 to 12 points (image 2 of 2)

松子 Songzi @_SongzioeveOriginally posted 2026.10.28

Open original on X

Artificial Analysis 刚把智能指数升到 v5，60% 改用私有测试集或不公开答案。结果最扎心的一条：开源和闭源的差距从 3 分拉到 12 分。已放权重的最高分是 DeepSeek V4.1 Flash 和 GLM-5.3，都是 32，还打不过 Claude Haiku 5.5 的 36。所以之前那 3 分，有多少是刷熟了公开题？

x.com/_songzioeve/status/2115271451259646319Archived 2026.10.28 12:02