AI 导读
OpenAI 或许已为 Codex 的限额找到一个合理的折中方案。 允许任务在额度降到 0% 后继续运行对用户友好，但容易被滥用（很多人可能已经读到过相关情况）。硬性停止能防止滥用，却会让正当工作无法完成。Luna Reserve 现在提供了一种独立的、有上限的兜底方案，在高级模型限额用尽后启用。 我在自己的套餐里还没找到它。
正文
OpenAI may have found a sensible middle ground for Codex limits.
Allowing tasks to run past 0% was user-friendly but exploitable (as many probably have read about). Hard stops prevented abuse but left legitimate work unfinished. Luna Reserve now offers a separate, capped fallback once the advanced-model limit is reached.
I couldnt find it in my plan yet.
来源：@kimmonismus · x.com