Primary sourcerohanpaul_ai
View sourceEvent ·Multi-source
SemiAnalysis 发布 GDPx 基准,Claude Opus 5.5 平均仅得 12%
SemiAnalysis 发布 GDPx 基准,测试 AI 智能体在 11 个领域的专家级工作,任务来自 14 家专业数据合作方。榜首 Anthropic 的 Claude Opus 5.5 平均得分只有 12%,Google 的 Gemini 4 Carbon 以 10.9% 居次。与 OpenAI 的 GDPval 直接在提示词中给出上下文不同,GDPx 要求智能体自行翻查工作区,包括一个 3,983 个文件的真实收购案复刻。剔除网络安全、生物和对齐三类任务后,GPT-6 Astra 的得分从 9.6% 升至 11.8%,超过 Opus。金融和法律任务在通过/不通过评分下所有模型都得 0 分。
Current read
SemiAnalysis 搞了个 GDPx 基准,让智能体真去翻几千个文件干活,结果最强的 Claude Opus 5.5 也才拿 12 分,金融法律题全军覆没。
11 sources64° AI heatLast updated 2026/11/2 12:45:11
Evidence
11 sourcesRelated source 1kimmonismus
View sourceRelated source 2Blognone
View sourceRelated source 3量子位
View sourceRelated source 4Decoder
View sourceRelated source 5IT之家
View sourceRelated source 6TestingCatalog
View sourceRelated source 7Tibor Blaho
View sourceRelated source 8Inquirer: Technology
View sourceRelated source 9The Information
View sourceRelated source 10VTV Công nghệ
View sourceConflict 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.