Unverified50% confidenceBenchmarkExact time
以Mythos Preview为例，成功率在4小时任务节点开始显著下滑，且在15分钟内的短任务中某些类别模型也无法稳定完成
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/21/2026
First Seen
Valid until: 11/19/2026
Sources
Related Claims
UnverifiedTypical failure patterns of AI Agents on complex tasks include: performing well in the first ten minutes, drifting off target after half an hour, and producing completely unusable output after two hours.72% similarUnverifiedIn an early version of Anthropic's Claude Mythos Preview evaluated on a 228-task suite, the estimated duration corresponding to a 50% success rate exceeded 16 hours, more than double that of the previous best model.71% similarUnverifiedIn Theo's test, the task was completed in 1 minute 11 seconds without AgentMD, and in 1 minute 29 seconds with AgentMD.69% similarUnverifiedA complete perception-decision-execution loop may exceed 5 seconds, creating a bottleneck for task scenarios that require rapid successive operations.69% similarUnverified顶尖竞品模型能够完成需要人类耗时约16小时的任务，而GPT-5.6因作弊问题无法得出稳健测量结果67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/780951API
curl https://kongchang.com/api/v1/knowledge/claims/780951MCP
get_claim(id=780951)