OpenAI Pro $200 plan reopens at half the API spend value
Tibo posted the bad news a day before the announcements, and Theo called it transparent and painful in the same breath.
OpenAI's @thsottiaux, posting as Tibo, said the $200 Pro subscription reopens to new subscribers with a changed usage calculation, and that if you do the math it nets out at half the dollar in API spend compared to the old Pro $200 plan. He posted it the day before OpenAI's announcements, ahead of what he described as lots of new exciting things coming to the subscriptions.
it will net out at half the dollar in API spend compared to the old Pro $200 plan
The post lays out four reasons. There is a commitment not to reintroduce the 5 hour limit, so the weekly usage can be spent whenever you want. There is a guarantee that over time subscribers always get more work done at an increasing level of quality, funded by models getting more efficient and those gains passed down as API price cuts. There is a stated refusal to inflate API list prices to make the subscription look generous, with GPT-6 Sol and GPT-6 Luna cited as this week's example, both introduced at 50% of their previous price. And there is a teaser that more things are being added to the subscription that will not draw on usage, unnamed for now.
The read for anyone pricing off a flat plan
The broader argument in the post is that the gap between a subscription dollar and an API dollar should keep shrinking until most people just buy usage as needed. That is a direction of travel worth noticing if your own cost model assumes a fixed monthly coding plan absorbs whatever your agents throw at it. Tibo's claim is that you still get more work done than on the Pro $200 plan of a month ago, but the per dollar allowance is explicitly moving.
Theo's response was to quote the half the dollar line back and note the timing, hoping this was the end of bad news and that Dev Day the next day was the start of good news. That is one reaction from one builder, not a measured impact, and nobody has run the new calculation against real workloads yet.
The thing that is actually unusual here is the sequencing. A vendor telling heavy users the ratio is getting worse before the keynote, rather than letting them discover it in a usage dashboard two weeks later, is rare enough that the loudest critic led with respect. Whether the promised additions that do not draw on usage make up the difference is the part nobody can price yet.
Mad respect for the transparency here but this hurts a lot