OpenAI is quietly testing a pay-to-reset quota feature. Here’s what we know.

A mysterious new function has been appearing for some users of ChatGPT’s Plus plan.

A Reddit user on ChatGPT’s $20 monthly plan posted in the ChatGPT forum on Monday, saying they were presented with a black “pay $8 to reset” button when they logged in and found that their weekly usage quota had been maxed out.

The pay-to-reset feature wasn’t formally announced before it was first spotted.

The company introduced banked Codex resets for eligible Plus and Pro users in June, including a free reset at launch and resets earned through referrals.

According to OpenAI’s Codex, a “full reset” is defined as renewing the user’s applicable weekly quota and eliminating the waiting period, typically 5 hours, before the user’s quota automatically renews.

The user could choose not to pay and simply wait it out.

The $8 option could be a significant saving for professional plans, which can run at about $25 to $50 a week.

An OpenAI spokesperson told Business Insider that the company is exploring ways for some users who hit their quotas to buy additional usage, part of its push to make its products both more accessible and functional.

A Codex reset immediately restores a user’s depleted AI usage back to 100%.

If someone triggers a manual banked reset, it clears their current weekly caps, allowing them to continue coding without waiting for the normal rolling schedule.

Their weekly reset date will move to roughly seven days after they redeem it.

The new function is another instance of OpenAI being unusually generous with freebies lately, from free usage-limit resets to free ChatGPT Plus for college students.

In the growth stage of their business, companies see freebies as investments in customer acquisition — they encourage heavier usage, which may become a habit, making it more likely users switch to premium tiers.

For OpenAI, freebies and unexpected promotions are also ways to encourage users to abandon their loyalty to its chief rival, Anthropic.

Anthropic is playing the same game, offering Claude users promotional credits and, in some cases, substantial free usage.

For both labs, the idea is to get people to adopt or switch their workflows and eventually spend around their ecosystem.

On Sunday, OpenAI’s head of Codex, Thibault Sottiaux, and Anthropic’s head of Claude Code, Boris Cherny, traded swipes at each other after a user mixed up the two labs and mistakenly tagged Sottiaux in a complaint about his Anthropic account.

When Sottiaux clarified that he did not work at Anthropic, Cherny replied, saying Anthropic is hiring if he would like to make the switch.

Sottiaux replied saying that GPT-5.6 Sol is “awesome.” “To celebrate this, together with the fact that I’m not going anywhere…

I have reset usage limits for all paid users of ChatGPT Work and Codex.

In a similar back-and-forth in May, OpenAI CEO Sam Altman and Claude’s developer account both said they were giving away expanded access to their tools. “Codex is the best AI coding product and we want to make it easy to try,” Altman posted on X.

He added that OpenAI would give businesses two months of free Codex usage if they signed up within the next 30 days.

Less than an hour later, Anthropic posted that it was increasing Claude Code’s weekly limits by 50% until July 13. “Live now for all Pro, Max, Team, and seat-based Enterprise users,” Anthropic wrote. “We’re excited to see what everyone builds!”