Artificial Analysis 更新模型能力指数，新增两项测试Chinese only
Artificial Analysis 发布 Intelligence Index v5，新增真实代码任务和终端科研任务测试；60% 的指数现基于私有测试集或未公开答案。其测试结果中，Claude Opus 5.5（max）以 58 分居首，GPT-6 Astra（max）以 57 分紧随其后。该机构还推出 Progress Index v1，用于比较不同发布时间模型的能力变化。
1 developments2 reports2 sources
Event developments 1
Artificial Analysis updates its model capability index with two new benchmarks
Artificial Analysis released Intelligence Index v5 with new benchmarks for real-world coding tasks and scientific research workflows in a terminal. Private test sets or held-out answers now underpin 60% of the index. In its results, Claude Opus 5.5 (max) leads with 58 points, followed by GPT-6 Astra (max) with 57. The organization also launched Progress Index v1 to compare model capability over time.
View source reports · 2 sources · 2 reports
All reports 2
Artificial Analysis releases Intelligence Index v5 with revised model evaluations
The new index places 60% of evaluation weight on private test sets or hidden answers. In the newly included FrontierCode 2.0, open-source repository maintainers judge whether code changes are worth merging. DeepSeek V4.1-Flash 1017 (max) scored 28%, the highest in that category among models with released weights, but its overall index score was 32, below the leader’s 58.
1 more source reported this
- Artificial Analysis 模型评测
Artificial Analysis updates its model capability index with two new benchmarks
Artificial Analysis released Intelligence Index v5 with new benchmarks for real-world coding tasks and scientific research workflows in a terminal. Private test sets or held-out answers now underpin 60% of the index. In its results, Claude Opus 5.5 (max) leads with 58 points, followed by GPT-6 Astra (max) with 57. The organization also launched Progress Index v1 to compare model capability over time.
1 more source reported this
- IT之家 科技新闻