OpenAI's new $500/month Pro 500 plan brings Astra Ultrafast speed to ChatGPT Work and Codex
OpenAI's Pro 500 plan costs $500/month, offers 25x the Plus usage allowance, and is the only tier with Astra Ultrafast access.
On 29 September 2026, at its DevDay event, OpenAI launched a new subscription tier: ChatGPT Pro 500, priced at $500 per month. It sits above Pro 100 ($100/month) and Pro 200 ($200/month) in what is now a three-tier Pro lineup, and it carries one exclusive that the other two tiers do not: access to Astra Ultrafast inside ChatGPT Work and Codex.
If you have been watching OpenAI’s pricing move upward over the past year, this is the tier that was circulating as “ChatGPT Pro Max” in leaks earlier in 2026. It shipped, more or less as expected.
What you get with Pro 500
The headline number from OpenAI’s DevDay recap is that Pro 500 includes roughly 25 times the usage allowance of ChatGPT Plus. Everything available on the lower Pro tiers comes along: Pro models, Codex for coding tasks, deep research, and image creation. The exclusive addition is Astra Ultrafast, a speed mode for GPT-6 Astra that is available only to Pro 500 subscribers (and eligible Enterprise and Edu plans) within ChatGPT Work and Codex. Buying additional credits on Pro 100 or Pro 200 does not unlock it.
It is worth being clear about what Ultrafast actually means in practice. In Codex, GPT-6 Astra Ultrafast generates tokens at up to 300 tokens per second, which OpenAI describes as roughly 8 times faster than Standard. That figure measures token generation speed specifically. A task that spends most of its time running tests, waiting on tools, or executing code will not finish 8 times sooner overall. The speed gain applies to the model’s output phase, not the full end-to-end task.
That speed comes at a cost against your allowance. Ultrafast draws on your included usage at 8 times the Standard rate. If you exhaust your included allowance and move to purchased credits, the rate drops to 6 times the Standard rate. OpenAI’s API pricing for Astra Ultrafast is $60 per million input tokens and $300 per million output tokens, compared to $10 and $50 for Standard. The model picker inside ChatGPT is where you select Ultrafast; it pulls from your included allowance first, then your credit balance once that is spent.
One practical restriction to note: Ultrafast is not available to workspaces that require inference residency outside the United States.
What this means for you
For most people, the honest answer is: Pro 500 is not for you, and that is fine.
If you are a developer or power user who runs Codex heavily throughout the day and has been hitting limits on Pro 200, the combination of a much larger included allowance and faster token generation could meaningfully change how you work. Tasks that currently stall while waiting on the model become faster, and you are less likely to burn through your allowance before the reset. For that kind of intensive, continuous use, $500/month may well be justifiable.
If you use ChatGPT for a mix of writing, research, and occasional coding, Pro 100 or Pro 200 almost certainly covers your needs. Ultrafast is a speed optimisation for high-volume Codex workflows, not a capability that unlocks new things you cannot otherwise do.
The Pro 200 situation
Alongside the Pro 500 launch, OpenAI also reopened Pro 200 to new subscribers. It had paused sign-ups on 10 September 2026 due to demand for the Astra model. New Pro 200 subscribers join on a lower included usage allowance than was offered previously.
Existing Pro 200 subscribers keep their previous allowance through 29 October 2026, after which they move to the new lower allowance. OpenAI is providing a one-time credit to help with that transition. Usage limits on these plans are firm: once you hit the limit, the model becomes unavailable until the allowance resets, and OpenAI support cannot manually raise or reset the limit.
The broader DevDay context
Pro 500 was not the only thing announced on 29 September 2026. OpenAI also gave Codex reusable cloud environments, automatic security scans for GitHub repositories, and a code review view in the desktop app. The Agents API gained Computer Use support, and a new Decisions API handles fast, single-decision tasks. GPT-6.1 Sol Ultrafast was flagged as coming soon. OpenAI also noted that ChatGPT now reaches 1.2 billion users per week.
For the Pro 500 announcement specifically, subscriptions are available immediately through ChatGPT on the web. The plan is billed monthly only; there is no annual billing option and no option to pay several months in advance.