發布日期：2026-07-18
你正要把長企劃、會議紀錄和程式碼一起交給 Codex 時，最在意的常是它到底記得住多少內容。這版更新了 GPT-5.6 Sol、Terra 與 Luna 隨附的操作說明，並把三者的 context window（單次對話能保留的內容容量）校正為 272,000 token。token 是 AI 處理文字的計量單位；把它想成可放進同一張工作桌的紙張數，容量標對了，你才不會拿錯桌子大小來安排材料。後兩項變更也把這份模型資料和提示詞修正帶回 0.144 版，範圍只鎖定 GPT-5.6，沒有宣告三個模型新增能力。
When you are about to hand Codex a long plan, meeting notes, and code together, the key question is how much it can keep in view. This release refreshes bundled guidance for GPT-5.6 Sol, Terra, and Luna, and corrects each model's context window—the amount of material one conversation can hold—to 272,000 tokens. A token is a unit AI uses to process text. Think of it as the number of sheets that fit on one worktable: label the table correctly, and you can plan your pile of material properly. The two follow-up changes also backport that model-metadata and prompt correction to 0.144, limited to GPT-5.6; they do not announce new abilities for these models.
Refreshed bundled instructions for GPT-5.6 Sol, Terra, and Luna, and corrected their context windows to 272,000 tokens. (#33972, #34009)
#33972 Backport refreshed bundled model metadata to 0.144 @sayan-oai
#34009 Narrow 0.144 hotfix to GPT-5.6 prompts and context @sayan-oai