动察 Beating AI Flash: OpenAI recently rolled out GPT-6 Astra in batches, sending Codex quota reset cards to Plus, Pro, and Business users who got access later.

However, some users noticed that the quota after using the card seems to deplete unusually fast. One user compared records of consuming 14% of the weekly quota before and after the reset. Before the reset, it took approximately 114 million input tokens and 738 model responses; after the reset, only 63.6 million tokens and 489 responses. Based on this, they estimated that the reset card actually only provides half of the normal weekly quota.

That said, this data set isn't a strictly controlled comparison. Both periods primarily ran Astra, but the reasoning tiers and Agent usage patterns differed. Codex head Tibo subsequently denied the claim outright, stating there is no difference in available quota before and after the reset.