Unverified85% confidenceBenchmarkExact time
Models like GPT-4 and Claude 3.5 surpassed 90% accuracy on MMLU, significantly diminishing its discriminative power
1
Sources
85%
Confidence
Medium-term (~90 days)
Relevance
8/3/2026
First Seen
Sources
Text Arena: A Deep Dive into the LLM Battle Evaluation Platform and Its Ranking Mechanism
Related Entities
Related Claims
UnverifiedOn 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.69% similarUnverifiedGLM 4.6's performance on code evaluation benchmarks such as HumanEval, MBPP, and SWE-bench, as well as comprehensive capability benchmarks such as MMLU and GPQA, is on par with Claude 3.4 and Claude 3.4.5.67% 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%67% similarUnverifiedGemini 3.5 Pro scored 89.4 on the MMLU Pro benchmark, leading GPT 5.5 and Claude65% similarUnverified在 GSM8K 基准上，标准思维链准确率约为 56%-63%（GPT-3 规模），PAL 可提升至 72% 以上65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/680092API
curl https://kongchang.com/api/v1/knowledge/claims/680092MCP
get_claim(id=680092)