精选理由
有用户说GPT-5.6干活更利索比之前版本强一大截，但Codex还不够好，来看看他的体验。
一位用户在X平台表示，经过一段时间使用后，认为GPT-5.6是自GPT-4相比GPT-3.5以来最明显的性能跳跃。该模型更善于按照指令完成任务，执行力显著增强。用户同时指出主要问题在于Codex（编程界面）仍需大量改进。
原文 · @koltregaskes
After using it for a while now, I have to say GPT-…
After using it for a while now, I have to say GPT-5.6 is the first model where I've clearly seen a jump in performance and capabilities over the previous version - since GPT-4 over GPT-3.5. I know we've been going up in dot releases, so it’s been hard to tell recently, but .6 was definitely a big jump.
GPT-5.6 just seems more capable - it does what you ask and gets on with the work. Most of my issues are with the harness (i.e. Codex), which still needs a lot of work.