As of November 15, 2026, the improvement is plausible, but it is not a universal quota increase. Some Pro plans received larger allowances. Plus did not. Separately, OpenAI introduced improvements to reusing the context of long-running conversations. A bigger fuel tank and an engine that wastes less fuel can feel similar from the driver's seat, but they are different changes.
The disappearing quota trick
You ask Codex to fix one line. The visible request is tiny. Behind the scenes, an agent may repeatedly process a large conversation, repository context, tool descriptions, and previous results. The instruction is a sticky note; the processing can resemble moving the entire filing cabinet. Context capacity, subscription allowance, and API token prices are distinct measurements.
What changed in October?
On October 20, OpenAI announced a selectable 12-hour prompt-cache setting for its API. Reusable input can be processed more efficiently and at a lower cached-input price. Around that time, an OpenAI team member also announced persistent reuse when ChatGPT-signed-in Codex and Work users resume old conversations. On October 23, a further announcement said changing reasoning effort mid-conversation should no longer invalidate that reuse. These changes offer a credible explanation for better endurance on repeated work, but the API setting does not itself increase a Plus subscription allowance. [2][3][7]
The implementation is not perfect proof of every user's experience. In an October 22 GitHub issue, one user reported zero cached input on the first request after an overnight break, followed by approximately 99.8% cached input on the next request. That is an individual report, not a platform-wide statistic. [6]
Who actually got more allowance?
OpenAI's official plan help says Pro 100 and Pro 500 received twice their earlier Work and Codex allowances; on Pro 200, the lower allowance doubled. Plus usage was unchanged. Existing eligible subscribers received the change automatically. A separate October report about incorrect allowance application to around 40,000 accounts concerned a correction for affected accounts, not a new benefit for everyone. [1]
What about the latest November updates?
The November 13 Codex CLI changelog lists interface improvements and bug fixes, not a newly announced doubling for all Plus accounts. OpenAI also says ChatGPT and Codex allowances are scheduled to become shared from December; today's usage pattern should not be projected indefinitely. [4][5]
How to test your own improvement
Compare similar work with the same model, reasoning effort, speed setting, and repository size. Check usage before and after, for both the five-hour and weekly windows. The CLI offers /status; other supported interfaces have a Usage view. Record whether a long thread was resumed, its context was compacted, or an initial request consumed unusually much. Without that baseline, “it lasts much longer” is an observation, not a measured percentage. [8]
The punchline: Pro customers may have a larger tank; repeat Codex users may have less waste. For Plus users, “everyone got twice the quota” is the wrong story. Perhaps the smartest improvement was getting the agent to stop rereading the entire office archive every time someone moved a comma.
References (1)
[1] OpenAI Help Center, About ChatGPT Pro tiers: https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers [2] OpenAI API, Changelog (2026-10-20): https://developers.openai.com/api/docs/changelog [3] OpenAI API, Prompt caching: https://developers.openai.com/api/docs/guides/prompt-caching [4] ChatGPT Learn, ChatGPT & Codex changelog: https://learn.chatgpt.com/docs/changelog [5] OpenAI Help Center, Understanding ChatGPT and Codex usage limits: https://help.openai.com/en/articles/20001625-understanding-chatgpt-and-codex-usage-limits [6] openai/codex issue #56050, user report, 2026-10-22 (not official measured general behavior): https://github.com/openai/codex/issues/56050 [7] Tibo (OpenAI), public announcement, 2026-10-20 US time: https://x.com/thsottiaux/status/2112712984644784534 [8] OpenAI Help Center, Using Codex with your ChatGPT plan: https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan [9] Tibo (OpenAI), account-limit mismatch correction, 2026-10-20 US time: https://x.com/thsottiaux/status/2112592264686149749 [10] Tibo (OpenAI), cache preserved on reasoning change, 2026-10-23 US time: https://x.com/thsottiaux/status/2113770573994299544
