Unverified50% confidenceBenchmarkExact time
On the WebArena browser agent benchmark, the then-strongest GPT-4 agent scored 14.41%, while humans scored 78.24%.
1
Sources
50%
Confidence
Long-term
Relevance
10/6/2026
First Seen
Sources
Related Entities
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%75% similarUnverifiedGPT-6 Astra peaked at 98% on Frontier Math Tier 4, and still reached 83% without enabling reasoning or using chain-of-thought.72% similarUnverifiedIn CAD World's 200-task evaluation, the best-performing GPT-5.4 achieved a success rate of only 17.5%.72% similarUnverifiedIn EvoArena testing, GPT-5.5, which achieved the highest Step Level score, had an accuracy of just over 50%, while Gemini-3.1 Pro and GLM-5.1, which performed best at the Chain Level, reached only approximately 28%–29%.71% 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%69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/981294API
curl https://kongchang.com/api/v1/knowledge/claims/981294MCP
get_claim(id=981294)