Codex Hits 5 Million Users, Chasing Claude Code

OpenAI's Codex coding agent has surpassed 5 million users, a milestone revealed not through an official press release, but in an informal post by lead engineer Tibo Sottiaux on X late on May 31. "Five million users would agree," Sottiaux wrote. "Resetting the limits tomorrow morning to celebrate. Time to go /fast." The post generated over 870,000 views within hours.

Resetting usage limits has become Codex’s signature celebration. OpenAI CEO Sam Altman established the pattern earlier this year: each time the AI agent surpasses another million-user threshold, fast-mode caps are temporarily lifted to allow unrestricted access. The 5-million-user mark triggers the fifth such reset, a ritual Altman has promised to continue until the platform reaches 10 million users.

Codex User Base Doubles in Two Months

The 10-million-user target now appears far more attainable than it did just a quarter ago. Codex began the year with approximately 2 million weekly active users. OpenAI reported that figure rose to 3 million by April 21, and surpassed 4 million by mid-May. Reaching 5 million just two weeks later represents a near-doubling of its user base in approximately two months.

This growth extends beyond user count. In February, OpenAI noted that overall Codex usage had doubled since mid-December, following an October report that daily usage had surged tenfold compared to early August. The reference to '/fast' in Sottiaux's announcement underscores this trajectory, as each limit reset serves as a critical stress test of the infrastructure's ability to handle massive concurrent traffic.

The Signal from 10,000 NVIDIA Employees

A substantial portion of this operational load originates from within the very chipmaker that powers the hardware. NVIDIA recently confirmed that over 10,000 of its employees across engineering, legal, finance, sales, and human resources are actively using Codex powered by GPT-5.5, OpenAI’s latest frontier model. CEO Jensen Huang sent a company-wide email urging comprehensive adoption, concluding with: "Let's jump to lightspeed." Employees have described the tool as "mind-blowing" and "life-changing," noting that complex debugging tasks that previously required days are now resolved in hours.

This acclaim extends beyond NVIDIA's internal teams. Siqi Chen, founder of Runway, evaluated GPT-5.5 on Codex's 'xhigh' configuration and asserted that it significantly outperforms Anthropic's models for coding agents, a point underscored by this head-to-head matchup. Given Anthropic’s historical dominance in the developer space, this is a significant assertion. The industry reception suggests that Codex is no longer merely competitive for OpenAI, but is actively challenging Claude Code, the established market leader.

Claude Code's Formidable Market Lead

However, the incumbent remains a formidable competitor, and the competitive gap remains substantial. On May 28, Anthropic announced that Claude Code's annualized run-rate revenue has surpassed $2.5 billion—more than doubling since the beginning of 2026. Additionally, weekly active users have doubled since January 1, while enterprise subscriptions have quadrupled. A recent analysis estimates that Claude Code now generates 4% of all public commits on GitHub, doubling its market share in just one month.

Technical benchmarks reinforce this competitive dynamic. On the SWE-Bench Pro evaluation, GPT-5.5 achieved a score of 58.6%, compared to 64.3% for Claude Opus 4.7, demonstrating that Claude Code maintains its lead in resolving the most complex engineering tasks. Consequently, Codex’s competitive advantage currently lies in its rapid adoption and expansive reach: 5 million users, broad enterprise deployment at NVIDIA, and an engineering lead who routinely expands compute access to celebrate milestones. Whether this momentum will ultimately disrupt Anthropic's financial dominance in the coding agent market remains a question for future resets to answer.