安全测试者批 GPT-6 Sol 太「高效」、分析太浅

Soso fun ✨ @Soso_fun_yt原推发布于 2026.09.23

在 X 上打开原推

My opinion on GPT-6 Sol after testing it: useless for cyber, and too "efficient". GPT-6 Sol tends to finish within the first few minutes, even on tasks that clearly require much deeper exploration, and its findings are very superficial. Even 5.6 Sol goes deeper in its analysis ! 😅 Hope that it will be fixed, this problem was also on Astra when they had nerfed it. This urgently needs to be fixed. 🙏🤍

这条是引用着另一条写的：「OpenAI 发布 GPT-6 Sol 与 Luna，API 降价 50%」