OpenAI has added a $500-per-month ChatGPT Pro tier for intensive users and heavy workflows, and it includes dots, OpenAI’s always-online cloud agents. The $200-per-month Pro plan is open to new subscribers again, but with a revised usage allowance worth roughly half of its former API-equivalent value. Existing subscribers temporarily retain legacy limits and transition credits. The change is often described as a halving, yet that shorthand combines different measures: an overall usage comparison, a GPT-6 Pro message cap and the removal of a short-term rate limit. They do not affect every task in the same way.
Three limits that should not be confused
OpenAI reopened Pro sign-ups after pausing them because of capacity constraints. Alongside that change, it revised how plan usage and token consumption are calculated. Tokens are the units an AI model processes when it reads input and produces output.
Worked out against the previous plan, the revised allowance delivers roughly half the earlier API-equivalent usage value. An API, or application programming interface, lets software use a model directly; an API-equivalent comparison estimates what similar usage would cost through that route. It does not mean a Pro subscriber receives $100 in API credit or pays only $100. A breakdown of the revised terms using Anthropic’s Claude points the same way, putting the plan’s usage multiplier at over 10x, down from over 20x.
| Measure | Earlier arrangement | Revised arrangement |
|---|---|---|
| Reported Pro usage multiplier | Over 20x | Over 10x |
| GPT-6 Pro messages | 200 per window | 100 per window |
| Five-hour rolling rate limit | Applied previously | Will not be reinstated |
The GPT-6 Pro figure is a message limit for that model, not a count of all tasks a subscriber can run. The usage multiplier, by contrast, describes the plan’s overall allowance, so the two figures should not be read as the same limit.
▲ Flexible use of a weekly allowance
The missing five-hour limit also matters. Subscribers can use their entire weekly allocation in one five-hour session if a project requires it, rather than being stopped by the former rolling rate limit. That flexibility does not increase the weekly allocation. Someone who concentrates work into a single session may benefit from it; someone who needs more total usage across the week may still encounter the lower allowance.
What new and existing subscribers face
A new subscriber choosing the reopened $200 plan enters under the revised usage structure. An existing subscriber is temporarily keeping legacy limits and receiving transition credits. Those protections make the immediate effect different for the two groups, but the available details do not set out how long the temporary arrangement lasts or how much a particular subscriber’s credits cover.
The stated direction is to increase Pro subscribers’ output and quality over time, with model efficiency gains passed along through lower API prices. OpenAI also plans to add features that will not count against weekly allowances. Those are plans and commitments, not a replacement for checking the limits that apply to a workload now.
Why “half” is not a universal price change
The revised plan looks less favorable on the reported overall allowance, and GPT-6 Pro’s message cap falls by half. Other comparisons have a different result. GPT-6 Sol and GPT-6 Luna have API prices roughly 50% below those of earlier model iterations. For workloads centered on lighter models, lower underlying prices may offset much of the allowance change. The reported assessment puts such users at roughly break-even under the new structure.
Astra, the resource-intensive flagship model, presents a different trade-off: its base pricing did not fall while plan limits changed. Heavy Astra use may therefore cost more in effective terms than lighter-model use. In effect, the savings on lighter models are balanced by a premium on top-tier frontier compute.
Where the $500 tier and dots fit
The $500 Pro tier is built for intensive users and heavy workflows, and it includes dots, cloud agents that stay online and keep running tasks in the background. Since dots run on Astra, keeping agents active around the clock may cost considerably more as time goes on. Cloud agents are also spreading across competing AI companies, which makes regular cost monitoring part of running them.
▲ Different workloads, different limits
These comparisons explain why neither “everything was cut in half” nor “nothing changed because model prices fell” captures the whole update. The allowance estimate describes one way to compare plan value. The 200-to-100 figure describes one model’s message cap. API price cuts describe the cost of particular models, the five-hour change describes when weekly usage can be spent, and the $500 tier adds a separate option for heavier, agent-driven work.
What to check before choosing a plan
Start with the work you actually do. Separate lighter-model tasks from GPT-6 Pro messages and resource-intensive agent runs, then compare each against its relevant limit. If your work arrives in occasional bursts, the ability to spend a weekly allocation in one session may matter more than a rolling cap did. If you regularly use the full weekly allowance, the lower reported allocation may matter more.
Existing subscribers should also check which legacy limits and transition credits still apply to their accounts before comparing plans. New subscribers should assess the revised $200 plan rather than assume they will receive the temporary protections. For a workflow that does not need a fixed subscription, direct API usage or a local model setup is another cost comparison to make, not a guaranteed cheaper choice.
The practical conclusion is to compare limits by workflow, not by a single headline percentage. Check your model mix, message use, agent needs and weekly spending pattern first; then decide whether the revised $200 plan, the $500 tier or another way to access models fits the tasks you need to complete.