OpenAI cancels Pro 200 cut, doubles ChatGPT Pro Codex limits

The cut is off. On October 29, one day before the Pro 200 reduction was due, OpenAI posted in its ChatGPT release notes that it canceled the cut and doubled the ChatGPT Pro Codex limits on Pro 100 and Pro 500. Good news if you run Codex on Pro seats. But look at the pattern: your included AI quota changed twice in 30 days, and both changes came from the vendor.

What actually happened

On September 29, at DevDay, OpenAI told Pro 200 subscribers by email that their Work and Codex allowance would drop on October 30. We covered that cut the same day. Reports put the drop at 20× to 10× the Plus allowance.

The October 29 release note reverses it and goes further:

• Pro 100 and Pro 500 now include twice their previous Work and Codex allowance.
• The Pro 200 reduction is canceled. Subscribers who still had the higher allowance keep it.
• Pro 200 subscribers on the lower allowance get twice that allowance from October 29, which brings them up to the same limit.
• Business Premium also gets more Work and Codex usage. The note gives no multiplier for it.

The note covers Work and Codex usage only. It says nothing about chat message caps, Plus, or the one-time usage credits that came with the original cut. Don't assume any of those changed. Check your own usage page.

Why ChatGPT Pro Codex limits matter for your business

A reversal is still volatility. In September the smart move was to downgrade seats or move work off Pro 200. In October that same move would have left headroom on the table. If your team re-planned staffing or tooling around the cut, you planned around a number that lasted one month. That will happen again, in both directions.

Measure what you use, not what the plan says. The allowance is a ceiling, not a forecast. Log Codex tasks per developer per week, and note who actually hits the limit. In most teams we audit, two or three people hit the ceiling and the rest never get near it. A doubled allowance only matters for those two or three.

Don't build a production job on a subscription seat. Seat allowances can move with one email. Anything that runs on a schedule, like nightly test generation, bulk refactors, or ticket triage, belongs on metered API keys with a budget cap and a fallback model. Then a quota change is a line on an invoice, not an outage.

Sources: OpenAI ChatGPT release notes, The Next Web.