{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreiaqunutwyu34j736wsxqpvxq6i3r5icqxehspdh6y2emcnevceymy",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mnvqteh73so2"
  },
  "path": "/t/codex-rate-limits-discussion-thread/1378553?page=19#post_382",
  "publishedAt": "2026-06-10T02:01:48.000Z",
  "site": "https://community.openai.com",
  "textContent": "Update on Case #09851164.\n\nI want to clarify the core issue because this is not simply about whether 500M tokens is “a lot” in absolute terms. The issue is the relative quota accounting on the same Pro 20x account, in the same current billing cycle, compared with usage just days earlier.\n\nMy weekly reset is June 13, so the current weekly window is roughly June 6 to June 13. In this window, my user-side observed usage was approximately:\n\n  * June 7: less than 100M tokens\n  * June 8: about 100M tokens\n  * June 9: about 390M tokens\n  * By June 10: roughly 500M-700M cumulative tokens, but the weekly quota was already around / above 80% consumed.\n\n\n\nThat ratio does not make sense.\n\nWithin the same current monthly Pro billing cycle that started June 1, I had much heavier usage before this weekly window:\n\n  * June 1: about 450M tokens\n  * June 2: about 200M tokens\n  * June 3: about 200M tokens\n  * June 4: about 500M tokens\n  * June 5: about 100M tokens\n  * June 6: about 200M tokens\n\n\n\nAnd immediately before that:\n\n  * May 31: about 880M tokens\n  * May 30: about 840M tokens\n  * May 29: about 300M tokens\n  * May 28: about 180M tokens\n\n\n\nI did not see the same weekly-limit collapse under much heavier usage. This is why I believe the problem is specifically with the current June 6-June 13 weekly window, not a generic historical complaint.\n\nIf OpenAI changed policy or ended a temporary double-credit / promo around June 5 or June 6, Support needs to say clearly whether that change applies to a paid ChatGPT Pro 20x / $200-month account. I am not relying on temporary free credits; I am paying for Pro.\n\nWhat I am asking OpenAI to investigate:\n\n  1. Are cached input tokens being charged against weekly quota at an abnormal/full weight?\n  2. Did the long-running-session / compaction cache-hit issue return?\n  3. Is Fast / Go mode being classified into the wrong quota bucket?\n  4. Was my account moved into a lower limit tier, rollout cohort, or hidden risk-control bucket?\n  5. Is the UI mixing general Codex weekly quota with the GPT-5.3-Codex-Spark bucket incorrectly?\n\n\n\nThis is incredibly frustrating as a paying Pro user. A generic answer like “limits may vary” does not explain why the same account can use far more tokens days earlier without this behavior, then suddenly burn most of the weekly quota with far less usage.\n\n中文补充：\n\n我再说清楚一点，这不是在争论 5 亿 token 绝对值大不大。问题是同一个 Pro 20x 账号、同一个 6 月账期里，6 月 6 日到 6 月 13 日这个周窗口的扣额度倍率明显不对。\n\n我的周额度重置时间是 6 月 13 日，所以这一周大概从 6 月 6 日开始算。本周我看到的使用量大概是：6 月 7 日不到 1 亿，6 月 8 日约 1 亿，6 月 9 日约 3.9 亿，到 6 月 10 日累计也就是 5-7 亿左右，但周额度已经烧到 80% 左右甚至更高。\n\n可是同一个月账期里，6 月 1 日到 6 月 6 日我用得更重：6 月 1 日约 4.5 亿，6 月 2 日约 2 亿，6 月 3 日约 2 亿，6 月 4 日约 5 亿，6 月 5 日约 1 亿，6 月 6 日约 2 亿。更早的 5 月 30 日、5 月 31 日甚至分别接近 8.4 亿和 8.8 亿，也没有出现这种额度崩掉的情况。\n\n所以这不是普通的“未来限额可能变化”，而是这一周的 Codex quota accounting 看起来明显有问题。请 OpenAI / Codex quota team 查后台，而不是只给通用政策解释。",
  "title": "Codex Rate Limits Discussion Thread"
}