Codex reaches 7M users as coding agents accelerate

Codex and ChatGPT Work reached 7M active users after a rapid post-launch surge around GPT-5.6 Sol. A July 12 update said the products hit 6M users in the prior 48 hours, followed by a July 13 update announcing 7M users and a banked reset for accounts. Earlier disclosures put Codex at 2M users in March and around 550k-700k users on Jan 1, supporting an estimate of around 10x user growth year to date.

The comparison with Claude Code remains difficult because its last cited update was roughly 2M users and $2.5B ARR in Feb. Claude Code’s relative silence may reflect a shift of coding activity to Claude Tag, where usage patterns are harder to compare with a CLI tool, but Codex’s recent growth has become a major marker in coding-agent adoption.

OpenAI also addressed GPT-5.6 Sol usage issues, including roughly 10% more usage from inference optimizations, a context limit rollback from 372k to 272k, reasoning-effort changes, and fixes for overactive multi-agent behavior. Broader agent coverage centered on Prime Intellect’s verifiers v1, cost-per-task competition, real-world agent benchmarks, vLLM support for Transformers models, Grok Build privacy concerns, and new research directions in continual learning and multimodal systems.