OpenAI cut the ChatGPT Pro $200 plan's usage from 20x to 10x and added a $500 tier, with old allowances honored until 29 October 2026: Reveneau

Why it mattersA team relying on a Pro $200 seat to power an unattended coding or research workflow has one month before the seat's effective throughput halves at the same price, so the choice is to move to $500, drop to $100, or shift the workload to the API.

A subscription tier that a lot of teams have built unattended coding and research workflows around just changed underneath them. The New Stack's Frederic Lardinois reported on 29 September 2026 that OpenAI has cut the ChatGPT Pro $200 plan's usage allowance in half and added a new $500 tier, with the numbers now published in OpenAI's own ChatGPT Pro help article. Existing $200 subscribers keep the old allowance until 29 October 2026, then drop to the new allowance at the same price.

The new ladder

Pro $20 remains the baseline at 1x. Pro $100 is 5x. Pro $200 is 10x, down from the previous 20x. The new Pro $500 tier is 25x and adds an "ultrafast" mode that runs at 8x the speed of the standard tier at 6x the token cost. OpenAI's own reason for the change, quoted in the help article: "New subscriptions that aren't eligible for grandfathering include a lower usage allowance than previously offered with Pro 200 to reflect our increasingly efficient models." Discussion on Hacker News, where the change went to the front page today, called the "ultrafast" price the tell: paying 6x per token to move 8x faster means the $500 seat drains at more than half the token spend of a raw API budget for the same work.

Who this hits and when

The 29 October 2026 grandfathering cutoff is the load-bearing date. For a solo developer running Codex from a Pro $200 seat, the seat's effective allowance halves in a month at the same price. For a team that stood up an internal agent or research runner on the same seat, the same happens. New Pro $200 subscribers do not get the grandfathering at all: they start on the new 10x allowance today. OpenAI publishes the tiers as multiples of the $20 baseline and does not publish the token or task counts each multiple works out to, which Hacker News commenters described as "hand-wavey" and lacking transparency. The multiples themselves are OpenAI's, from OpenAI's own help doc, and are the only quantitative statement OpenAI is making about what any of these plans includes.

What a team can actually do about it

Three real paths. Stay on the grandfathered $200 seat through 29 October and use the month to measure how close the workload sits to the new 10x cap, one week of logs is usually enough. If the workload was already close to the old 20x limit, moving the same work to the API on the current price sheet is a straightforward calculation and gives a monthly bill that scales with actual use rather than with a seat multiple that OpenAI can change again. Moving up to the new $500 tier is honest only when the workload was already running well over the old 20x, because 25x is not 2.5x what a real Pro $200 user was using; the new "ultrafast" mode at 6x token cost is the tier that will bite anyone who does not model the token draw before switching. The workflow to keep off the seat entirely is anything with a hard SLA, since a seat OpenAI can retune under a "reflect our increasingly efficient models" note is not a foundation for anything that has to run.

Source

The New Stack, OpenAI halves $200 plan allowance, launches $500 plan, by Frederic Lardinois, 2026-09-29.
OpenAI, About ChatGPT Pro tiers, help article, retrieved 2026-09-29.
Hacker News, ChatGPT Pro 500, community discussion, 2026-09-29.

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.