💻 OpenAI changed Astra's quota consumption
OpenAI's Codex and ChatGPT lead Tibo Sottiaux announced on X on September 6 that for heavy users, GPT-6 Astra's ChatGPT subscription quota consumption may become up to 3–4 times less, while model quality remains unchanged. The change came amid complaints: after the full banked reset on September 5, Plus users, according to their reports, hit the limit in about 20 minutes, while Pro users burned through a significant part of their weekly quota in a day.
🌍 The case shows the mechanics of frontier-model launch economics: Astra is significantly more expensive in inference than GPT-5.x, but it is included in the same fixed subscriptions. OpenAI recalculates request consumption against the quota rather than changing nominal limits — the model's "price" within a subscription turns out to be a flexible parameter that is adjusted as real traffic reveals the cost per request.
👤 If you use ChatGPT Plus, Pro, or Business, around 01:00 UTC on September 8, Astra's limits were reset for the second time, and long heavy sessions — agentic tasks, 3D modeling, and similar — now, according to OpenAI, consume up to 3–4 times less quota. Practical step: measure your actual consumption in the coming days to understand the new norm.
Source 1: https://x.com/thsottiaux/status/2096717905614524491
Source 2: https://x.com/thsottiaux/status/2097043464538264003