ChatGPT Pro Stays at $200 a Month but Loses Half Its Allowance as OpenAI Launches the $500 Pro 500 Tier

The loudest reaction at OpenAI's DevDay wasn't to Dots or the new models — it was to a price sheet. OpenAI has relaunched its $200-per-month ChatGPT Pro subscription to new subscribers while cutting its included usage roughly in half, and simultaneously introduced a $500-per-month Pro 500 tier that takes over as the plan with the highest allowance and exclusive access to the fastest model mode.

The cuts are specific. Pro 200's usage allowance in ChatGPT Work and Codex drops from 20 times the $20 Plus plan to 10 times. GPT-6 Pro messages in Chat fall from 200 per week to 100. Existing subscribers keep their current limits through October 29; after that they receive a one-time grant of 62,500 usage credits — which OpenAI values at $2,500 — that expires on December 31, 2026. The one mercy: the old 5-hour rolling cap is not coming back, so the weekly budget can be spent at any pace.

OpenAI's product and platform lead Tibo Sottiaux previewed the change on X the day before the announcement, and did the math himself: under the new structure, the dollar value of included API spend on Pro 200 nets out at roughly half of the old plan. His defense is that subscribers will "still get more work done than if you were on the Pro $200 subscription one month ago," because the models inside the plan got cheaper and faster — GPT-6 Sol and Luna API prices were cut by about 50%, and the new GPT-6.1 Sol delivers near-Astra quality at about a fifth of Astra's price.

The new ladder is transparent to the point of blunt: Plus at $20 for 1x usage, Pro 100 at $100 for 5x, Pro 200 at $200 for 10x, Pro 500 at $500 for 25x. Every unit of Plus-equivalent allowance costs $20 at every tier — the volume discount that once justified the flagship plan is gone. What Pro 500 buys beyond raw quota is speed: exclusive access to Ultrafast mode for GPT-6 Astra, generating tokens up to 8 times faster in Codex, with an Ultrafast tier for GPT-6.1 Sol promised soon.

Developer reaction was immediate and negative, heavy on promises to migrate: "Ok guys, back to Claude." The sharpest complaint is the flagship problem — OpenAI cut prices on the models few Pro subscribers were buying it for. Astra's API price is unchanged, so for users who chose Pro specifically for the strongest model, the same $200 now buys half as much of exactly what they signed up for. As Chinese coverage of the keynote put it, the options are pay more or accept less.

There is competitive context, too. The 20x allowance at the top tier had been the industry standard: Google's $200 AI Ultra plan and Anthropic's top $200 plan both still offer it, and Google's comes with storage and other perks. If OpenAI's restructuring sticks, it will be the first frontier lab to break the 20x floor — and the test case for whether subscribers follow.

The deeper shift is in how OpenAI frames what you're buying. The company is moving away from all-you-can-use token buckets toward something closer to hiring: pricing tied to delivered work and how fast it runs, with speed itself a paid tier. It is a rational structure for a company reportedly raising $30 billion at a $1.4 trillion valuation and burning compute commitments measured in hundreds of billions — and a blunt message that the era of subsidized abundance at the top of the consumer ladder is over.