deepseek-v4.1POST
用户质疑 Artificial Analysis 评分低估 DeepSeek 1017
该用户认为 1017 的 45 分与实际表现不符，有时能打赢 55 分的 K3.1，两者应同列 1.5 档。
Beamsters
@Beamsters15 天前收录原推发布于 2026.10.25
Beamsters @Beamsters1原推发布于 2026.10.25
I think at this point Artificial Analysis should improve how it determines intelligent score. Deepseek 1017's 45 is much closer to the frontier and sometimes even beat K3.1's 55. Both should be tier 1.5 breakaway from those mid packs that cluttered around 40-45.
x.com/beamsters1/status/2114182729915043890采集于 2026.10.25 12:01