OpenAI said it has resolved issues that caused some users of its coding agent, Codex, to hit usage limits faster than normal.
Thibault Sottiaux, the engineering lead for OpenAI’s Codex, said in a late Monday X post that the coding tool was doing more work behind the scenes than intended, which in turn consumed more compute.
Over the weekend, Codex users took to X to flag that they were reaching the limit faster compared to just a week ago for similar tasks. In response, Sottiaux said OpenAI set up a Sunday “warroom” to investigate and rolled out an across-the-board reset of user caps.
Usage limits track how much compute an AI coding task has used. Codex’s usage limits are displayed on a user’s dashboard as a percentage. More intense tasks burn through credits faster, and the limit varies by subscription tier.
Sottiaux said in his Monday post that features like auto-review, which checks lines of code without human intervention, and helper “subagents” were sometimes running more often than they should, running twice, or retrying too aggressively after errors.
He added that the Codex dashboard was also erroneously showing activity that hadn’t actually been charged to users.
“All fixes are now deployed, and we’ve added more detailed monitoring so we can detect background-usage regressions sooner. We’ll continue watching the results closely,” Sottiaux wrote, adding that OpenAI had “fully reset” usage limits again for all users.
what has happened to codex limits?
they went from practically infinite tokens to weekly quota gone in 2-3 days on the pro 20x plan.
— Alexander Hoff ❍ (@axhoff) June 28, 2026
Coding is one of the more computationally-intensive tasks for AI, meaning incorrect tracking could rack up expensive bills.
“It really seems something is off. I have the $200 plan, and I had to work hard the entire week to burn through the 7 days’ usage,” wrote a software engineer called Adam in a Sunday X post. “The last two days burned the entire week’s usage in one day each, and I had to use reset for the 1st time.”
Companies like OpenAI and Anthropic are betting big on coding because it’s one of the clearest examples of people willing to pay for AI. Any issues relating to billing and downtime are sensitive for those companies because they risk customers switching to a different service.
“Still skeptical. I stopped using codex for a few days now, because the depletion is brutal,” wrote Srinivas Pendela, a developer, in a Tuesday X post.
OpenAI’s Codex limit issue follows an outage earlier this month. Anthropic has had similar problems, and a Claude outage in March underscored how much software engineers now rely on the technology, as some bemoaned having to write code by hand again.
Codex usage limits will be fully reset again in the next hour and we will credit one additional reset into your bank for your own usage over the next 24 hours.
We investigated reports that Codex usage was being consumed faster than expected. There wasn’t one central issue, but a… https://t.co/UdCSbW0BGs
— Tibo (@thsottiaux) June 29, 2026
At the same time, AI companies have gradually been reducing usage limits as the soaring popularity of their tools has strained compute resources. In March, Anthropic adjusted Claude usage caps during peak hours. Some software engineers have been restructuring their days while they wait for usage limits to reset.
More broadly, AI providers have been moving away from all-you-can-eat use of their tools, while companies have started limiting employee AI use to keep costs down.
Update, June 30: This story was updated to reflect that OpenAI said it rolled out a fix for Codex issues.
Read the full article here
