OpenAI has launched a $500-a-month Pro plan for ChatGPT, giving subscribers 25 times the allowance of the $20 Plus tier in the U.S. and exclusive access to an Ultrafast speed option. The pricing, however, stays linear: on a per-unit basis, Pro 500 costs the same as Plus, and faster output burns through allowance at a higher rate.
According to OpenAI’s DevDay 2026 official video, Pro 500 was introduced in San Francisco on Sept. 29. The plan is available only as a monthly subscription, which puts the annual cost at $6,000.
U.S. pricing tiers show no bulk discount at the top end
OpenAI’s official Pro plan description lists the following U.S. tiers:
- Plus: $20 for 1x allowance
- Pro 100: 5x allowance
- Pro 200: 10x allowance
- Pro 500: 25x allowance
From Plus to Pro 500, the monthly fee rises by 25 times and the allowance rises by 25 times as well. That leaves the price per 1x allowance at $20 in both cases. The report’s point is simple: Pro 500 does not introduce a volume discount.
Pro 200 used to stand out. It previously came with 20x allowance, which worked out to $10 per 1x. That earlier "buy more, pay less" structure is now gone.
The reset began in September after GPT-6 Astra demand surged
The report traces the change back to Sept. 10. GPT-6 Astra went live during the week of Sept. 3, demand jumped, and Pro 200 became the most compute-intensive subscription in the lineup. OpenAI then paused new subscriptions and blocked upgrades from Free, Go, Plus, and Pro 100.
Thibault Sottiaux, who leads Codex at OpenAI, described the demand wave as unprecedented in an interview with Fortune.
Pro 200 reopened on Sept. 30, but with lower included usage. Work and Codex allowances were reduced from 20x Plus to 10x, while GPT-6 Pro weekly messages were cut from 200 to 100. Those changes take effect on Oct. 30.
Existing Pro 200 subscribers can keep the old allowance through Oct. 29. They will also receive a one-time $2,500 usage credit that expires at the end of the year.
Ultrafast is exclusive to Pro 500, but speed comes with a direct allowance tradeoff
Ultrafast is the signature feature of Pro 500 within the Pro family. In Codex, it pushes GPT-6 Astra output to as much as 300 tokens per second, about 8 times standard speed.
On the API side, speed improves by about 6 times, and pricing scales in the same direction. Ultrafast Astra costs $60 per million input tokens and $300 per million output tokens, versus $10 input and $50 output for the standard version.
The subscription logic follows the same pattern. Within the included allowance, Ultrafast deducts usage at 8 times the standard rate, while Fast mode deducts at 2.5 times. Once the included allowance is exhausted, purchased credits are charged at a 6x rate.
Using that math, a Pro 500 subscriber who spends the full 25x allowance entirely on Ultrafast gets the equivalent of about 3.1x Plus at standard speed. That is less than the 5x standard-speed allowance included in Pro 100.
Put differently, Pro 100 costs $100 a month and provides 5x allowance at standard speed. Pro 500 costs five times as much, but if Ultrafast is used throughout, the total amount of work completed can end up lower.
Price and allowance scale in a straight line
The report argues that Pro 500’s monthly fee and allowance are both 2.5 times the reduced Pro 200 tier. In that framing, Ultrafast is a fair but not discounted trade: users exchange allowance for speed on a one-for-one basis. Faster output means less work completed under the same allowance ceiling.
Who the $500 plan is built for
The article says ChatGPT has 1.2 billion weekly users, while Work and Codex together have more than 35 million weekly developers. Pro 500 is aimed at the narrow top slice of that group.
For users who regularly exhaust the new Pro 200 allowance every week, buy extra usage on top of that, and treat waiting time as a real operating cost, the combination of 25x allowance and Ultrafast may be worth considering.
For most users, the report says Plus or Pro 100 paired with GPT-6.1 Sol may be enough. GPT-6.1 Sol was announced at the same event and is described as close to GPT-6 Astra in coding and computer-use performance. Its API pricing is about one-fifth of Astra’s: $2 per million input tokens and $10 per million output tokens. It is available in Codex and Work for Plus subscribers and above.
Rivals still cap personal plans at $200
The report also compares OpenAI’s new top tier with competing consumer plans. Anthropic offers Claude Max 5x at $100 and 20x at $200. Google AI Ultra ranges from $99.99 for 5x to $199.99 for 20x, with the higher tier having been cut from $249.99 at I/O in May this year.
After halving Pro 200’s allowance and adding a new $500 tier above it, OpenAI has become the only one of the three companies to push a personal subscription plan to that price level.