OpenAI’s Codex was launched at a time when Claude Code had already cornered most of the coding market, but it seems to have made a serious dent in the space.

Tibo Sottiaux, Codex’s product head at OpenAI, announced on X on May 31, 2026 that Codex has crossed 5 million users — and to celebrate, OpenAI is resetting usage limits the following morning. The post came as a reply to developer Siqi Chen, who had noted that switching back to a competing model after a few days on Codex reminded him just how much better the product had gotten. “It’s really not close,” Chen wrote.

From Research Preview to 5 Million Users in Just Over a Year

Codex launched as a research preview in May 2025, built on codex-1, a version of OpenAI’s o3 model tuned for software engineering. It could write features, fix bugs, answer questions about codebases, and propose pull requests — all running in isolated cloud sandboxes. The tool rolled out first to paid ChatGPT tiers before opening to Plus users in June 2025.

Growth was immediate. Within weeks of launch, CEO Sam Altman reported usage was up 10x. By October 2025, when Codex reached general availability, OpenAI noted that daily usage had grown more than 10x since early August — and that GPT-5-Codex had served over 40 trillion tokens in just three weeks.

The trajectory since then has been steep. Codex cleared 1.6 million weekly active users by early March 2026. By April 21, OpenAI confirmed the number had crossed 4 million weekly developers — a 6.7x jump since January. Now it’s at 5 million, with limits being reset to keep momentum going.

Enterprise Is Driving the Engine

The growth story isn’t just consumer developers tinkering on side projects. Enterprise adoption has been a central pillar of Codex’s rise. OpenAI says companies including Cisco, Nvidia, Ramp, Rakuten, and Harvey have deployed Codex across their developer teams. Cisco reportedly cut code review times by 50% and compressed project timelines from weeks to days.

Within ChatGPT Business and Enterprise tiers specifically, Codex users grew 6x between January and April 2026 — a signal that the real money is in organizational deployment, not individual subscriptions. OpenAI has also expanded Codex’s reach with a Slack integration, a Codex SDK for embedding the agent in custom workflows, and admin dashboards for enterprise oversight.

The npm download numbers underscore the breadth of adoption: from 82,000 downloads in April 2025 to over 14 million in March 2026 — 177x growth in under a year.

The Competitor It’s Chasing

Codex isn’t operating in a vacuum. Claude Code, Anthropic’s terminal-native coding agent, has a substantial head start. Launched in February 2025, Claude Code reached a $500 million run-rate within months and crossed $2.5 billion in annualized revenue by February 2026. Between 41% and 68% of developers surveyed use Claude or Claude Code, and the tool accounts for roughly 4% of all public GitHub commits worldwide.

Weekly active Claude Code users doubled between January 1 and February 12, 2026 alone. Dario Amodei has noted that the average Claude Code developer spends 20 hours a week with the tool — a level of engagement that goes well beyond casual use.

Anthropic’s lead in coding benchmarks has also been durable. Claude models have topped LLM coding leaderboards for 18 consecutive months — a streak that began with Claude Sonnet 3.5 in June 2024. OpenAI has been chipping away at that lead, and the benchmark battles between the two companies have become an almost daily occurrence.

The Broader Context

The AI coding boom is straining infrastructure across the industry. GitHub recorded 1 billion commits in all of 2025; it’s now on pace for 14 billion in 2026. AI-agent pull requests jumped from 4 million in September 2025 to 17 million by March 2026. The market Codex is entering is not just large — it is expanding at a pace that makes room for multiple winners.

That context matters when reading Codex’s 5 million user milestone. It’s an impressive number, and the rate of growth is real. But Claude Code’s enterprise footprint, revenue lead, and developer engagement depth mean OpenAI still has ground to make up. Sottiaux’s celebratory post is a signal that OpenAI is treating this as a genuine strategic fight — not a side experiment.

The limits reset tomorrow morning. The competition, though, doesn’t.