Primary sourceArtificial Analysis
View sourceEvent ·Multi-source
MLCR-AA 排行榜更新:Claude Sonnet 5.5 与 GPT-6 Astra 并列第一
Artificial Analysis 更新了基于 Wisedocs 医疗长上下文推理基准 MLCR 的排行榜 MLCR-AA,改为将模型作为病历审阅者进行提示,并由三模型评审团打分。Claude Sonnet 5.5(max with fallback)与 GPT-6 Astra(max)以 41.7% 并列第一,GPT-6.1 Sol(max)为 38.3%。阿里巴巴 Qwen3.8 2.4T A95B 是最强开源权重模型,得分 30.6%。完整性方面 Claude 模型包揽前 18 名,Claude Sonnet 5.5 达 64.8%;准确性方面 OpenAI 模型占优,GPT-6.1 Sol(max)达 96.7%。无任何模型通过率超过 42%,中位数模型通过率低于 15%。
Current read
病历审阅榜单更新了,Anthropic 家模型答得最完整,OpenAI 家最不容易瞎编,两大阵营各有强项,开源的 Qwen 也上榜了。
11 sources63° AI heatLast updated 2026/10/21 15:53:56
Evidence
11 sourcesRelated source 1量子位
View sourceRelated source 2宝玉
View sourceRelated source 3lmarena.ai
View sourceRelated source 4TestingCatalog
View sourceRelated source 5arXiv cs.AI
View sourceRelated source 6The Rundown AI
View sourceRelated source 7elvis
View sourceRelated source 8AIGCLINK
View sourceRelated source 9Tech in Asia
View sourceRelated source 10@koltregaskes
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.