OpenAI doubles ChatGPT Pro limits and cancels the planned Pro 200 cut, a day before it was due to bite
A day before existing Pro 200 subscribers were due to drop from 20x to 10x the Plus allowance, OpenAI reversed course: usage limits are doubling across Pro and Business Premium, and the cut is off. Plus subscribers, notably, get nothing yet.
- Core vocabulary: a task is one test with defined inputs and success criteria; a trial is one attempt; a grader is logic that scores some aspect; the transcript is the complete trial record; the outcome is the final environment state.
- Grade transcript vs. outcome: a booking agent may announce 'your flight is booked,' but the outcome is whether a reservation actually exists in the database — most serious suites grade both.
- Sandboxes: each trial starts from a clean, isolated environment; shared state creates correlated failures and can inflate scores (one internal eval's model gained an advantage by reading git history left by previous trials).
- Graders: code-based (fast, cheap, reproducible, brittle); model-based LLM judges (flexible, non-deterministic, need calibration); human (gold standard, expensive, doesn't scale). Key lessons: grade what the agent produced, not the path — and calibrate LLM judges against humans, giving the judge an 'Unknown' way out.
- Success criteria: capability evals start at a low pass rate; regression evals sit near 100%. Use pass@k (at least one success in k attempts) vs. pass^k (all k succeed) deliberately, and always write a reference solution that passes every grader — one frontier model scored 42% until researchers fixed the eval, then 95%.
One day before the deadline, OpenAI blinked. On October 29, Thibault Sottiaux, the OpenAI lead for Codex and ChatGPT Work, announced that usage limits for every Pro tier and Business Premium would double immediately — and that the planned cut to the $200 Pro plan was dead. For subscribers who had spent a month bracing for half the service at the same price, it was a last-minute reprieve.
The reversal matters beyond one tier. Pro subscribers are OpenAI's heaviest users — developers running long coding sessions and agentic work through Codex and ChatGPT Work. When a company halves what its most committed customers get, then doubles it back under pressure, it shows where the leverage sits: with power users who can walk.
The cut that never landed#
The story starts on September 29, at DevDay, when OpenAI redrew its Pro ladder. Alongside a new $500 Pro 500 tier, it reopened the $200 Pro 200 plan with a much smaller allowance: in Codex and Work, from 20 times the ChatGPT Plus baseline to 10 times, and the weekly GPT-6 Pro chat allowance reportedly from 200 messages to 100. The $200 price did not move.
Existing subscribers got a grace period rather than an instant cut: anyone holding an active Pro 200 subscription during the late-September eligibility window kept the old allowance through October 29, then was due to step down on October 30. As compensation, OpenAI offered a one-time grant reported at 62,500 usage credits, valued at about $2,500 and expiring December 31, 2026. The arithmetic fuelled the anger: a grant expiring in two months in exchange for a permanent monthly reduction.
By October 28, with the deadline under three days away, frustration boiled over in public. A widely shared post framed the change as a permanent cut papered over by a one-time payment, and it spread fast among developers who live inside Codex daily. OpenAI's framing — that cheaper, more efficient models would still let subscribers get more done — never landed with people who had bought the plan precisely for its generous ceiling.
What actually doubled — and what did not#
Sottiaux's announcement covers the shared Work and Codex allowance on Pro 100, Pro 200 and Pro 500, plus Business Premium. OpenAI had heard subscribers' feedback and would not reduce Pro 200 limits, he said, so existing users' usage stays where it was — and on top of that, limits double across the tiers, effective the same day with no action required.
| Plan | Price / month | What changed October 29 |
|---|---|---|
| Pro 100 | $100 | Work and Codex limits doubled |
| Pro 200 | $200 | Planned cut to 10x cancelled; existing allowance kept, and limits doubled |
| Pro 500 | $500 | Work and Codex limits doubled; Ultrafast remains exclusive to this tier |
| Business Premium | — | Limits doubled |
Two things stayed put. Dots, OpenAI's always-on assistants, keep their extended limits, and Dot usage remains separate from a subscriber's default allowance — a deliberate carve-out, since a Dot launching Codex jobs on your behalf would otherwise burn your own quota. And OpenAI has not published exact new numerical limits per tier, describing the change only in relative terms.
Two questions remain. First, the fate of the 62,500-credit grant: with the cut cancelled, it is unclear whether affected subscribers still receive it. Second, Plus. Early coverage notes the $20 Plus tier sees no change, leaving its subscribers — the largest paying group — watching Pro users get a doubling they did not.
Why subscribers revolted#
The backlash was never really about one ratio. Pro 200's original appeal was that it bought usage at a discount: $200 for twenty Plus-sized allowances, roughly $10 each. The redrawn ladder priced every tier at the same $20 per Plus-equivalent, turning Pro from a volume deal into a simple prepayment. For casual users that distinction is invisible; for developers running Codex agents for hours a day, it was the whole product.
Subscribers have seen this pattern before. In July, Anthropic trimmed Claude's top-tier access similarly — cutting bonus usage, keeping the flagship model only at reduced levels for some plans, and handing out a one-time credit. Both companies hit the same wall: demand for their most capable models outran all-you-can-eat subscription economics. This time, OpenAI's customers pushed back loudly and early enough to get the decision reversed before it took effect.
What to watch#
Watch three things. First, whether OpenAI's Help Center gains concrete numbers — vague doublings are hard to plan around. Second, whether the credit grant survives the reversal or quietly disappears. Third, whether Plus gets anything at all. If the next announcement extends relief downward, this episode becomes a precedent: complain loudly, in public, before the deadline — and the terms can still move.
For now, the outcome is simple. The cut scheduled for October 30 will not happen, Pro limits are higher than yesterday, and subscribers who organized against the change got what they asked for — a rarer ending than it should be in the subscription economy.
Sources: OpenAI's October 29 announcement as reported by BleepingComputer and The Hype, with context from BigGo Finance. The 62,500-credit figure is as reported by The New Stack via secondary coverage, unconfirmed in OpenAI's Help Center. Follow-up to our September 29 coverage, OpenAI launches $500/month Pro 500 tier. Images are AI-generated editorial illustrations.