爆料：OpenAI 内部急于推出 Opus 5.5 级模型

Token Gremlin @TokenGremlin原推发布于 2026.09.27

在 X 上打开原推

A lot of you probably don’t know this, but there’s a pretty clear consensus inside OpenAI that they need to ship something at Opus 5.5 level, and FAST. Something smaller than GPT-6 Astra, cheaper than or at least priced like GPT-6 Sol, available across Chat / Work / Codex, with solid usage limits for Go, Plus, and Pro, the incredible personality of GPT-4o, and the unmatched professionalism of GPT-4.5. Easy, right? Honestly, that’s not even a very long wishlist considering the level of engineering I’m watching happen in real time. They are absolutely capable of pulling this off.