Xiao Huang: How capable is GPT-5.2 in specific tasks like programming and mathematics?
DOORM: GPT-5.2 Thinking achieved an 80% score in the programming benchmark SWE-bench Verified and a 100% score in the mathematics competition AIME 2025. This demonstrates its strong capabilities in complex reasoning tasks, even surpassing the previous generation of specialized vertical models
.
发表回复
要发表评论，您必须先登录。