OpenAI is testing a new model for Codex called "GPT-5.2-Codex-Max," and it's already rolling out to users with a subscription.

Some users have spotted a new model, GPT-5.2-Codex-Max, when they ask Codex what model it is using.

OpenAI rolled out Codex with GPT-5.2 in December, but at that point, no "max" variant was offered to paid users.

When OpenAI announced GPT-5.2-Codex, the company confirmed that its agentic tool can stay on track for long tasks, keep a large repo’s context usable through compaction, and handle heavy changes like refactors and migrations without losing the plot.

Codex 5.2 also seems more reliable with tool use, works better on Windows workflows, and its stronger vision means it can understand screenshots, UI bugs, and diagrams you share during a coding session.

If you look at how GPT-5.1-Codex-Max sat above the regular versions and delivered another jump, the pattern suggests a 5.2 “Max” could be the next bump again.

OpenAI could share details of GPT-5.2-Codex-Max in the coming days.