Event ·Multi-source

Artificial Analysis 发布 Decision Index v2,Drex Max v1.1 以 73 分领先

Artificial Analysis 发布 Decision Index v2,私有数据集占比从 25% 提升到 45%。新增 Decision Score 指标,同时奖励正确性与校准置信度,并加入 AA-Paperclip(1500 份文档,占 30%)和 AA-Symmetry(120 道变体问题,占 15%)两个基准。Nace AI 的 Drex Max v1.1 以 73 分领先,GLiDE v1.2 与 Mammoth 27B 分别为 73 和 72 分。成本差异显著:每 100 万任务的费用从 Celeris-1.1 Decision 的 14 美元到 Clef 的 859 美元相差超过 60 倍,Claude Haiku 5.5 (Max) 和 GPT-6 Luna (Max) 得分 77 和 76,但成本高出 15 倍以上。

Current read

决策模型排行榜 v2 出炉,Drex Max、GLiDE、Mammoth 咬得死紧,选模型前先看看这份榜单和成本对比。

5 sources42° AI heatLast updated 2026/11/8 23:51:02

Evidence

5 sources
Primary sourceArtificial Analysis
View source
Related source 2Aran Komatsuzaki (论文推介)
View source
Related source 3TestingCatalog
View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.