Unverified50% confidenceBenchmarkExact time
MMLU涵盖57个学科的多项选择题，GPT-4在此测试上首次突破86%的人类专家水平
1
Sources
50%
Confidence
Long-term
Relevance
7/15/2026
First Seen
Sources
Related Claims
Unverified在GPT-4发布初期，主流模型在GPQA上的准确率约为35-40%，人类领域专家基准线约为65%，随机猜测为25%65% similarUnverifiedModels like GPT-4 and Claude 3.5 surpassed 90% accuracy on MMLU, significantly diminishing its discriminative power63% similarUnverifiedClaude Opus 5在智能体搜索BrowseComp上得分30.8%，与GPT系列基本持平；工具增强多学科推理64.7%；生物学人类可解问题上高达90.1%63% similarUnverifiedOn mainstream benchmarks such as MMLU and HumanEval, the score gaps among Claude 3.5, GPT-4o, Gemini 1.5 Pro, and Grok-2 have been compressed to within the range of statistical error.61% similarUnverifiedOn the Agent's Last Exam benchmark, GPT-5.6 Soul (extra high tier) scored nearly 54%, while Fable (max tier) scored 45%.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/521285API
curl https://kongchang.com/api/v1/knowledge/claims/521285MCP
get_claim(id=521285)