Expired60% confidenceBenchmarkExact time
Stanford University's AgentBench benchmark test in 2024 showed that GPT-4-level models have an end-to-end success rate below 30% on complex tasks requiring cross-system coordination.
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
7/12/2026
First Seen
Valid until: 10/10/2026(expired)
Sources
Zuckerberg Admits AI Progress Falls Short of Expectations: The Industry Truth Behind the Layoffs
bilibili小溪_YouTube搬运笔记本7/8/2026
Related Claims
VerifiedStanford University once published a research report documenting performance fluctuations of GPT-4 on specific tasks across different time periods.79% similarVerifiedA 2023 Stanford University study found that GPT-4's performance on certain tasks showed a measurable decline over several months.78% similarUnverifiedResearch in 2024 showed that models such as GPT-4 and Claude, assisted by auxiliary tools, are already capable of independently solving approximately 30%–40% of medium-difficulty CTF challenges.78% similarUnverifiedWhen OpenAI's API endpoint names remain unchanged, the underlying model weights may be silently updated; a 2023 joint study by Stanford and UC Berkeley found significant performance differences in GPT-4's responses to the same prompts across different time periods, with accuracy on certain tasks dropping sharply from 97% to 2%.77% similarUnverifiedA 2024 study from UC Berkeley found that in complex multi-step tasks, GPT-4-level models exhibited tool-call hallucinations at a rate as high as 12%–18%.77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/489815API
curl https://kongchang.com/api/v1/knowledge/claims/489815MCP
get_claim(id=489815)