Codex reaches 4 million weekly users and resets rate limits

• OpenAI said Codex passed 4 million weekly users on April 21, up from more than 3 million in early April, according to OpenAIDevs' announcement and the linked OpenAI post.
• The company paired that milestone with another usage-cap change, with Thomas Sottiaux's post saying rate limits would reset again within hours and Sam Altman's post saying the reset would hit all plans that day.
• The official framing is not just consumer growth. The OpenAI post says OpenAI is launching Codex Labs and working with global systems integrators to roll Codex into thousands of engineering organizations, a point also visible in reach_vb's screenshot of the post.
• One day earlier, Thomas Sottiaux's incident update said Codex had a roughly 10 minute disruption, that Codex helped with remediation, and that a human found the key insight.

You can read the full enterprise rollout post, see the headline and rollout language in reach_vb's screenshot, and the day's other tell was operational: Thomas Sottiaux and Sam Altman both tied the 4 million milestone to yet another rate-limit reset.

OpenAI's public number moved fast. In early April, the company said Codex had more than 3 million weekly users. By April 21, OpenAIDevs and Sam Altman were both posting that Codex had crossed 4 million weekly users.

That is more than a headline bump. Thomas Sottiaux explicitly framed it as adding over 1 million users in less than two weeks, while the ARK chart reposted by jxnlco plotted the broader run-up as roughly 20x year to date, using OpenAI's public disclosures as inputs.

The official post attached to the milestone makes the bigger claim. OpenAI's announcement says OpenAI is launching Codex Labs and partnering with top GSIs to bring Codex to thousands of engineering organizations.

The same screenshot in reach_vb's post also quotes OpenAI saying enterprises are moving quickly to put Codex into real workflows across engineering and beyond. That shifts the story from raw usage growth to deployment motion inside larger orgs.

A smaller product detail surfaced later in the day. Thomas Sottiaux's follow-up said a new image model had become the default inside Codex, and that Codex could use it as another tool in its toolkit for website building and game asset generation.

The capacity story was visible before the 4 million announcement. On April 20, Thomas Sottiaux said a Codex incident had been mitigated after about 10 minutes, added that Codex itself helped with remediation, and said a rate-limit reset was incoming.

By the next day, the reset had become part of the milestone messaging. Thomas Sottiaux's 4 million post promised another reset within hours, and Wes Roth's earlier post had already amplified that OpenAI was resetting Codex rate limits across plans.

Taken together, the sequence is pretty clear: OpenAI is celebrating growth, expanding enterprise sales motion, and still actively tuning the guardrails around a service that had a short operational wobble less than 24 hours earlier.