OpenAI resets Codex and ChatGPT Work limits after 10M milestone
OpenAI's Codex lead announced a new paid-user usage reset for Codex and ChatGPT Work after a 10M milestone, with the reset expected to land within an hour of the July 21 post.
Paid Codex and ChatGPT Work users received another usage-limit reset after OpenAI's Codex lead posted a 10M-user milestone on July 21, 2026. The announcement said the reset would land within the next hour, making this a short-window access update rather than a normal product release.
The useful part is narrow: this is for paid users of Codex and ChatGPT Work. It does not say that every ChatGPT feature, every free account, or every regional plan received a broader entitlement change.
The Codex reset followed a 10M signal
Tibo Sottiaux posted that Codex and ChatGPT Work had reached "10M" and that a new usage reset for paid users would land within an hour. OpenAI's own Forum profile identifies Sottiaux as a Member of Technical Staff at OpenAI who leads Codex, so the post is a first-party operator signal, not a generic community rumor.
The independent Codex Resets tracker also captured the same July 21 announcement and listed it as the latest reset roughly 19 hours later during this scan. That tracker is not affiliated with OpenAI, so it is useful corroboration for timing, but the exact entitlement still comes from OpenAI's own plan surfaces and in-product banners.
What paid users should check now
OpenAI's Codex plan guidance says Codex, ChatGPT Work, ChatGPT for Excel, and Workspace Agents can draw from the same agentic usage and credit pool when those products are available on a user's plan. The amount consumed depends on task size, session length, selected model, and where the task runs.
That matters because a reset does not make usage unlimited. It replenishes the relevant allowance for eligible paid users, while longer repository work, extended agent sessions, and higher-capability models can still consume the pool quickly. Users near the limit should check the Codex usage page or limit banner for their account-specific options, including credits, upgrades, an available banked reset, or waiting for the next scheduled reset.
No universal expiry was published
The July 21 post gave a near-term activation window, not a detailed public policy page with a new end date, plan-by-plan table, or regional carveout. The safest read is therefore practical and limited: eligible paid users should have seen the reset shortly after the post, but anyone planning a large task should verify the meter inside Codex before starting.
This is especially relevant for teams using Codex or ChatGPT Work for long-running migrations, multi-file refactors, spreadsheets, research deliverables, or connected-tool workflows. Those jobs can fail awkwardly if the allowance is lower than expected halfway through a run.
Evidence and limits
The primary evidence is Sottiaux's July 21 X post, verified against OpenAI's own Forum bio for his Codex role. OpenAI's current Codex Help Center article provides the standing usage-pool context and explains that limits vary by plan. Codex Resets independently mirrored the reset announcement and timing.
This scan found no new official Help Center article that turns the July 21 reset into a permanent higher allowance. Treat it as a time-sensitive paid-user reset, then confirm the actual meter in your account.