Codex now offers more flexible pricing for teams

OpenAI quietly tweaked how businesses pay for Codex, its AI coding assistant, and the change is more meaningful than the one-line blog post suggests. Business and Enterprise customers on ChatGPT can now choose pay-as-you-go pricing instead of being locked into whatever seat-based plan they originally signed up for.

This matters because coding tools have a rollout problem that most SaaS products don't. A company might want fifty engineers testing Codex this month and five actually using it daily by next quarter, or the reverse, depending on how a pilot goes. Flat licensing forces finance teams to guess. Usage-based billing lets IT departments turn the dial without renegotiating a contract every time adoption shifts.

OpenAI isn't inventing this model — Anthropic and Google have leaned on consumption pricing for their own enterprise coding tools for a while now. But the timing says something about where OpenAI thinks the market is. Codex has been competing directly with GitHub Copilot and Cursor for engineering budgets, and pricing friction is often the difference between a team running a two-week trial and a team walking away before they even start.

What OpenAI didn't spell out is the actual rate card, or how pay-as-you-go pricing compares cost-wise once usage climbs past what a flat plan would have charged. That's the detail every CFO evaluating this will want before signing off, and it's conspicuously missing from the announcement.