Unverified50% confidenceBenchmarkExact time
On the HLE benchmark, COMED improved GPT-5.5's performance from 23.1% to 28.1%.
1
Sources
50%
Confidence
Long-term
Relevance
9/24/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn 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%.76% similarUnverifiedGPT-5.6 set a new all-time high score of 53.6 on the Agent's Last Exam benchmark, surpassing Claude Opus 4.5 by 13.1 points.72% similarUnverified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%72% similarUnverifiedIn the initial publicly released SWE-bench results, GPT-4's resolution rate was only approximately 1.7%, Devin claimed to reach 13.86%, and SWE-agent reached 12.5%.72% similarUnverifiedIn SWE-bench testing, the early GPT-4 resolution rate was below 2%, while Claude 3.5 Sonnet broke through 49%.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/946138API
curl https://kongchang.com/api/v1/knowledge/claims/946138MCP
get_claim(id=946138)