All Paid ChatGPT Accounts Will Undergo Global Reset Tomorrow at 10 AM Pacific Time
Thibault Sottiaux, head of OpenAI's Codex and ChatGPT, announced that all paid ChatGPT accounts will undergo a global reset tomorrow at 10 AM Pacific Time. He also apologized for the slow start of GPT-6.1 Sol, stating that the model experienced a massive load spike in the first two days but has now returned to expected speeds.
Earlier the same day, he wrote that GPT-6.1 Sol is one of the most in-demand models on both the API and subscription sides. Both ChatGPT and Codex were under heavy load at that time, and the team has increased capacity, with speeds expected to improve significantly within hours, approaching nearly double the speeds provided the previous day. The new post corrected "nearly double" to indicate that speeds have returned to expected levels.
The reset applies to paid accounts, and the post did not mention free tiers. It also did not specify whether it would clear the 5-hour window, weekly limits, or issue a new redeemable reset voucher. When GPT-6 Sol and Luna were released on September 22, he provided a banked reset for Plus, Pro, and Business accounts and stated that API prices would be permanently reduced by 50%. This time, the term used is "global reset," specifically for all paid ChatGPT accounts.
On July 12, he had previously issued a redeemable reset to 500,000 ChatGPT Work and Codex users, stating that the first such reset would be given to all Work and Codex users the next day in celebration of reaching 7 million active users. At that time, there was a malfunction where less than 10% of users who pressed reset did not see it take effect, with a window of about 2 hours, leading the team to issue another reset to everyone who pressed the button during that period.
Earlier on October 1, he had also posted that the main quota was nearly unlimited, so no resets could be issued, and a new solution needed to be found quickly. Hours later, the announcement changed to a global reset at 10 AM tomorrow. Reading both posts together, the reset is not a price reduction for the model but a replenishment of the paid quota consumed in the first two days.
Paid users are those who have already hit their limits for Plus, Pro, and Business accounts, while OpenAI sells computing time. This is a compensation driven by load spikes, not a monthly fee reduction. Usage from the first two days, which was depleted, will be restored to paid accounts. Those who benefited are the ones who fully utilized GPT-6.1 Sol in the first two days, while those under pressure are those who must endure slow speeds and limits until 10 AM tomorrow. The free tier is not included in this announcement.
Source: Public Information
ABAB AI Insight
OpenAI's handling of paid quota walls has become a fixed action. In July, they issued redeemable resets to Work and Codex users and provided a supplementary reset for a small ineffective window. On September 22, when Sol and Luna launched, they issued another reset for Plus, Pro, and Business accounts while permanently halving API prices. The global reset on October 1 follows the same path: replenishing quotas after the model release's depletion, rather than solving it with reduced monthly fees.
The computing power path is very specific. Sottiaux first stated that capacity had been increased, with speeds expected to be nearly double that of the previous day, and hours later announced a return to expected speeds, setting the reset for 10 AM Pacific Time. Money was not converted into new subscription tiers but rather reopening already sold paid time slots. The motivation is to preserve the user experience of GPT-6.1 Sol as the most in-demand model on both subscription and API sides, preventing the slow speeds of the first two days from becoming a reason for cancellations.
A comparable approach is Anthropic's increase of the 5-hour quota on model release day, followed by session discounts to retain users in the new product. Both sides no longer pretend that limits do not exist but instead make resets part of the release process. The current position is at an expansion peak: Sol is described as the most in-demand tier in history, with capacity still chasing load, not tightening functionality.
Structurally, this represents a transfer of pricing power. The listed price remains the monthly fee, while the actual pricing mechanism is the 5-hour window, weekly limits, and redeemable resets. The longer a model can run tasks, the faster it consumes, and the peaks on release day must be smoothed out by resets. What users purchase is not unlimited access but rather the portion that is acknowledged and uniformly replenished by the official the next morning after depletion.
ABAB News · Cognitive Laws
Peaks on release day are ultimately counted in the reset order.
Monthly fees are the listed price, while quotas are the pricing mechanism.
The higher the demand for a model, the sooner it hits the capacity wall.