两家同日发榜，各自都赢

Omar @notomarsol原推发布于 2026.09.26

在 X 上打开原推

benchmarks matter less to me every month Opus 5.5 and GPT-6 Sol launched on the same day and both labs posted AutomationBench scores Anthropic's chart: Opus 5.5 40%, GPT-5.6 Sol 28.8% OpenAI's chart: GPT-6 Sol 33.2%, Opus 5 26.9% neither could test the other's new model yet, so both compared against the old one and both look like the winner if you want to know which is better for you, give both the same real task from your work and compare