OpenAI is reopening its $200-a-month Pro subscription to new users tomorrow. But it is not coming back unchanged. Tibo from the OpenAI team said that, if you map the new Pro tier directly to API pricing, the usage works out to about half of what the earlier version gave people.
But he said that should not be treated as a clean 50% cut in real output. His point was simple: the models are more efficient and more capable now, so subscribers may still finish more work than they could a month ago.
No return to the old five-hour cap
Tibo said OpenAI has no plan to bring back the old five-hour usage cap. Pro users will still be able to spread usage across an entire week instead of slamming into a limit inside one five-hour block.
He said the company does not want to control heavy usage with "more frequent time-window limits." So OpenAI is changing how total weekly usage gets counted. That gets rid of the old setup that capped how much someone could use every few hours, though it also cuts the total compute value available each week.
Why the allowance looks smaller
Tibo said OpenAI wants Pro's value tied to actual work done, not just some big API dollar number in a headline. He said the long-term aim is for the subscription to let users do more work, use better models, and get more real output for every dollar.
That case only works if models keep getting cheaper. And this week OpenAI rolled out GPT-6 Sol and GPT-6 Luna, while cutting their API prices by 50% versus the promotional pricing for the GPT-5.6 series. Right now, OpenAI lists GPT-6 Sol at $2 per million input tokens and $10 per million output tokens. GPT-6 Luna is listed at $0.1 for input and $0.5 for output per million tokens.
OpenAI also said those price cuts came from better caching and inference efficiency, not some temporary discount. As people discussed in the OpenAI Developer Community, if the new Pro usage is calculated using the newest API prices, then a 50% drop in the allowance on paper, in dollar terms, does not automatically mean token volume or finished tasks fall by that same amount.
OpenAI does not want inflated API list prices to make subscriptions look better
Tibo said OpenAI wants to avoid an ugly incentive: leaving API list prices artificially high so subscription tiers can appear extra appealing through discounts, credits, or bundled access.
Instead, he said, the company wants to keep pushing API prices down while making the models better. That suggests the gap between subscriptions and pay-as-you-go API access could keep narrowing over time.
Tibo went a step further. He said that if inference costs keep falling, API pricing could someday get cheap enough that most users just buy what they need rather than depend on subscriptions for steep discounts.
More features are coming tomorrow, and some will not consume usage
The flip side of the Pro usage change is that OpenAI plans to announce more subscriber-only features tomorrow. Tibo said, "Tomorrow we will add more things that will not consume usage."
He did not say what those additions are. Still, he said there will be "many big announcements" ahead and argued that the new features should keep the subscription attractive.
If more of Pro's value starts coming from product features that do not count against model usage, then OpenAI's ChatGPT subscription model may be moving away from simply selling bigger usage buckets and toward bundling a wider work environment, tools, agent capabilities, and services.
ChatGPT Work and Codex already include GPT-6 Sol and GPT-6 Luna. OpenAI is also adding browser actions, document handling, and connections to external tools. GPT-6 Sol and Luna were formally added to Work and Codex on Sept. 22.