OpenAI unveiled several flashy new features during its annual developer’s conference Tuesday, from the always-on “Dot” work agent to a new, cloud-based security system that keeps constant watch for cyber threats. But one of the biggest — and to many ChatGPT users, the most infuriating — headlines from DevDay wasn’t announced on stage.
The news centered around the $200-a-month ChatGPT Pro 200 plan, which the company had paused for new signups earlier this month due to demand for GPT-6 Astra, OpenAI’s most powerful and token-hungry model.
First, the good news: OpenAI said it was taking new signups for the Pro 200 plan again. The not-so-great news? The plan’s relative usage limits compared to the $20-a-month ChatGPT Plus plan were being greatly reduced.
Welcome to another edition of Prompt Mode, your weekly AI newsletter.
I’m your host, Ben Patterson. Each week on Prompt Mode, I’ll be serving up analysis of the AI trends that matter to everyday users like you and me. Stay tuned for practical AI tips, hands-on experiences with the latest AI tools, and–you guessed it–prompts to help you get the most out of your AI assistants.
Thanks for reading, and if you like what you see, just sign up right here.
Previously, the Pro 200 plan got you 20 times more usage than Plus plan, but according to a rough chart shared by OpenAI’s Thibault Sottiaux on X, the $200-a-month Pro plan will now only offer 10 times the usage limits of Plus, a 50-percent reduction. (Existing Pro subscribers are getting a brief grace period along with usage credits to smooth things over.)
Naturally, those explanations aren’t going over too well in the ChatGPT community, with some users cancelling their subscriptions in protest.
OK, so why is this happening? Sottauix, the Head of Core Products at OpenAI, served up a few different explanations, including the debut of new DevDay functionality that demand “a lot of compute,” while noting that OpenAI passed on the chance to “compromise in other ways,” including the reintroduction of a rolling five-hour usage limit.
More importantly, Sottauix pointed out the recent 50-percent API rate cut for two of its newer and more efficient models, GPT-6 Sol (which already has a successor in GPT-6.1 Sol) and GPT-6 Luna. (API rates are essentially pay-as-you-go, versus opting for a flat subscription rate.)
“We don’t want to put an incentive on ourselves to artificially inflate the API list prices to make it look like [subscribers] are getting a lot (and workaround it through discounts, etc),” Sottauix wrote. “Instead we want to continue to both rapidly reduce prices and increase capabilities of models on the API.”
The result for Pro 200 subscribers getting their usage limits chopped in half is essentially a wash, although as Sottauix promises, “we guarantee that over time you always get more work done and with an increasing level of quality … as a result of models getting more efficient.” Sottauix also hinted at the addition of “more things to the subscription that won’t draw on the usage,” which turned out to include chats with OpenAI’s new Dot assistants.
Still, the changes to OpenAI’s Pro 200 plan point to a trend we’ve been tracking for months: namely, the big AI providers endlessly tinkering with their flat-rate subscription plans to keep pace with ever-changing and increasingly powerful AI features, particularly the token-heavy agentic ones. Expect more of the same in the weeks and months to come.
More in AI this week
- Here’s more about ChatGPT’s Dots, which are geared less toward home use and shopping and more toward collaborating at the office. [PCWorld]
- Google finally took the wraps on its revamped frontier Gemini model, Gemini 4 Aragon. Google-provided benchmarks look impressive, but the new model is being limited to “trusted testers” before rolling out to the general public. [Google]
- Gemini Gems are going away, but Gemini skills — which were formerly reserved for paying Gemini Spark users — are replacing them, and they’ll now work for free users. [PCWorld]
- President Trump has announced a “morally binding” accord with AI leaders over safety, while announcing that artificial intelligence should henceforth be called “super intelligence.” I’ll just leave it at that. [Reuters]
- Bill Gates sat down for Ezra Klein’s podcast to expand on his AI worries. If you’re looking for reassurance about AI apocalypse fears, you won’t find much here. [New York Times]
- Meta’s Muse AI agent has run headlong into various privacy snafus, including one in which it sent the home address of a user (who had asked Muse to take charge of his Facebook Marketplace account) to a complete stranger. [The Verge]
Prompt of the week: The “don’t build until I say” prompt
Claude Sonnet 5.5 has arrived, and with it comes a prompt that could help keep any AI agent from leaping into action before making a plan first.
The prompt, which comes directly from Claude maker Anthropic, concerns “open-ended” requests, including those that don’t strictly define boundaries for what the AI should and shouldn’t do.
Specifically, the prompt can help keep Sonnet and other AI models from spinning up charts, videos, and other “artifacts” when all you really wanted was just a plan.
That’s all for now!
Thanks for reading the latest issue of Prompt Mode.
Want more next week? Don’t forget to sign up to start receiving this newsletter in your inbox.
