预测 GPT-6 Sol 与 Opus 5.2 的强弱分工
他预测 Opus 5.2 编码更强,而 Sol 会在 Computer Use、3D 等方面胜出,并成为 200 美元档可长期运行的便宜模型。
My prediction on GPT-6 Sol vs Opus 5.2 > Opus 5.2 will definitely be a better coder than GPT-6 Sol Claude models already have a huge lead in terms of code quality and frontend, Opus 5 just screwed up the behavior of the model and its ability to play nice with an existing codebase. Opus 5.2 will definitely fix that since they learnt from their mistakes in Fable 5.1 > Sol will mog Opus on pretty much everything else Computer Use, 3D and other advantages that Astra has over Fable will definitely pass down to Sol. We will finally get a more affordable model we can have running for almost 24/7 on the $200 plan. What do you think?