推文

@harden_369741 · 2026-10-12 10:11

OpenAI just made context 80% cheaper to think with. GPT-6.1 Sol's default context in Codex and Work went from 272k to 500k overnight. Max is 1M. And they're not charging extra for long context. Why I care: compactions are where my long sessions go to die. Half my context lost because the model decided to 'summarize' my three hours of debugging into two paragraphs of vibes. My take: context windows are becoming infrastructure, like bandwidth. Nobody brags about bandwidth anymore. In a year nobody will brag about context either, they'll just assume it's infinite. #AI #OpenAI #coding

曝光 34 · 评论 0 · 点赞 0 · 书签 0 · 曝光/时 6.721739122135437