ChatGPT Pro Quotas Doubled: OpenAI Cancels Planned Codex Cut and Expands All Tiers

Just one day after Codex users were scrambling to burn through their usage limits before an announced cut, OpenAI reversed course entirely. In a post early October 30, 2026, Codex lead Tibo announced that all ChatGPT Pro usage limits—Pro 100, Pro 200, Pro 500, and Business Premium—are doubling immediately. The planned reduction to Pro 200 is cancelled, and Work and Codex quotas remain unchanged.

Tibo's phrasing was deliberately dry: "We heard your kind feedback." In the OpenAI community, where complaints about Codex limits have been relentless, the word "kind" carried the opposite weight.

What Changed

• Pro 100, Pro 200, Pro 500: All usage limits doubled, effective immediately.
• Business Premium: Also doubled.
• Pro 200 cut: Cancelled. The previously announced reduction is withdrawn.
• Work and Codex: Limits stay as they were.
• Personal Agent (dot): Expanded quota continues, billed separately from the default allowance.

Tibo closed with the standard line: better models are coming, efficiency work continues, and "we will keep working on it."

From the Reset Company to the Token Company

Longtime Codex users know one word better than any other: reset. Bug fix? Reset. Community outcry? Reset. Promotion? Reset. After a billing glitch in late August and another in October, many users developed a reflex: when you hit the limit, don't panic—wait, and it might reset tomorrow.

The world's most closely watched AI company had effectively been reduced by its own user base to a single button. Tibo leaned into the joke, framing the shift as moving from The Reset Company—reactive, patching limits after the fact—to The Token Company: giving enough upfront so resets stop being the main lever.

The distinction matters. A reset is damage control. A doubled quota is a retention strategy.

Why Quota Matters More Than Benchmarks

For heavy Codex users, the tool is no longer a chat interface. An agent sits in the background reading code, fixing bugs, running tests, and opening pull requests while the developer sleeps. The monthly quota is effectively how many hours that digital employee is allowed to work.

A Pro 200 cut meant telling developers that their unpaid overnight workforce was about to have its hours slashed. Workflows are built around the allowance: project setup, prompts, daily rhythm, even sleep schedules adjust to how much Codex can consume before hitting the wall.

The migration cost to Claude Code on the other side is already close to zero. When AI coding models differ by a few benchmark points, users can't feel the gap. But they can feel every hit of a usage cap. Quota has become a competitive lever as hard as model capability itself.

The Ford Analogy

The move echoes Henry Ford's 1914 decision to raise the daily wage from $2.34 to $5—more than doubling it, to the shock of his peers. The result was a collapse in worker turnover, and those $5-a-day workers could now afford the Model T they were building ($490 at the time, about five months' wages).

OpenAI is not being charitable. Every extra token given out is another weld holding a developer's workflow onto Codex. The more you use it, the harder it is to leave; the compute cost eventually gets recovered in lock-in. The lesson from 1914 and from this week is the same: when labor—or agent compute—is scarce, the company that pays more wins the people.

What This Means for Developers

The immediate takeaway is practical: if you were rationing Codex usage ahead of the feared Pro 200 cut, that rationing is no longer necessary. Doubled limits across all Pro tiers give room to run longer agent sessions, larger test batches, and more overnight PR loops without watching the meter.

The longer signal is that OpenAI is treating quota as a front-line product decision, not a back-end cost-control lever. The "Reset Company" era—fixing complaints by restoring limits after the fact—appears to be giving way to a strategy of setting expectations high and keeping them there.

Whether that holds depends on compute economics, not community feedback. If inference costs rise or usage spikes beyond what the doubled quotas are designed to absorb, the pendulum could swing back. For now, the developers who were burning through limits at 1 a.m. can stop. The quota doubled.