What benchmarks are left where human performance still exceeds frontier AIs?
retained text
Simple bench, ARC-AGIs and a lot of knowledge benchmarks have already surpassed human level. What are the most major benchmarks where people still outperform frontier LLMs? Do you think we can say that we achieved AGI where there is no longer any benchmarks where people surpass LLMs? If not, what is your definition of AGI?
document fetched 2026-11-12 00:31 · canonical https://reddit.com/r/singularity/comments/1xtbj62/what_benchmarks_are_left_where_human_performance