OpenAI says Codex reached 3 million weekly users and will reset usage limits at every new million up to 10 million

According to an April 7 post on X from Tibo Sottiaux, Codex has now reached 3 million weekly users. He said the jump from 2 million to 3 million happened in a little under a month, and that OpenAI is resetting usage limits to mark the milestone. Later the same day, Sam Altman echoed the announcement, saying OpenAI will do the same reset at every additional million users until Codex reaches 10 million weekly users.

This is a growth and operating story rather than a model launch. OpenAI did not pair the milestone with a new Codex model, a new pricing tier, or a permanent public limits overhaul. Instead, the company tied product adoption directly to policy changes around access. That is notable because usage milestones are often kept as internal scoreboard metrics; here, OpenAI turned the number into a public-facing mechanism for keeping builders active on the platform.

Milestone-based usage resets are an unusual product lever

The cadence is what stands out. OpenAI is not just celebrating one round number. It is promising another usage-limit reset at each additional million users up to 10 million. Based on the posts, that looks less like a structural reform of the product's long-term limits and more like a milestone-based release valve during a rapid growth phase. The company did not explain the exact per-plan changes, the duration of each reset, or the specific scope of the relief. Even so, the signal is clear: OpenAI is treating Codex growth as something that should immediately feed back into product policy.

The announcement matters because it suggests Codex is becoming a repeat-use workflow rather than a one-time curiosity. Moving from 2 million to 3 million weekly users in under a month points to very fast habit formation among builders. By foregrounding usage and retention instead of a flashy new feature, OpenAI is effectively saying the current problem is not awareness but sustained access at scale. Sources: Tibo Sottiaux's X post and Sam Altman's X post.