Skip to main content

OpenAI to Reset Codex Quotas Tomorrow, Fixes Usage Bug

OpenAI is taking swift action to address a wave of user complaints about unexpectedly fast usage consumption in Codex. Thibault Sottiaux, who leads the Codex & ChatGPT team, announced on social media that all subscribed users will see their usage quotas reset uniformly around 2 PM Pacific Time tomorrow.

The company's investigation into the abnormal usage surge uncovered three specific technical culprits. First, when users frequently included images in long sessions and the system underwent multiple context compressions, the underlying operational efficiency took a significant hit. Second, the resource usage of the Computer History feature remained unusually high at the P95 percentile and above. Additionally, the routine feature that automatically generates conversation titles consumed more resources than expected.

To tackle these issues head-on, the technical team has been working around the clock on fixes, with the optimization measures expected to go live tomorrow. But that's not all—during their debugging efforts, the team stumbled upon an entirely unrelated efficiency improvement. They plan to roll that out next week, aiming to further polish the overall user experience.

For users who've been frustrated by the rapid drain on their quotas, this reset comes as a welcome relief. The fixes should also help prevent similar issues from cropping up in the future. As OpenAI continues to refine its tools, it's clear that user feedback plays a crucial role in shaping these updates.

If you've been affected by the usage bug, mark your calendar for tomorrow's reset. And keep an eye out for the upcoming efficiency boost—it might just make your coding sessions smoother than ever.

Key Points

  • Quota Reset: All Codex subscribers will have their usage quotas reset tomorrow around 2 PM PT.
  • Root Causes Identified: Three technical issues—image handling in long sessions, high resource usage in Computer History, and overconsumption by the title generation feature—were behind the problem.
  • Fixes Rolling Out: Optimization measures are scheduled to go live tomorrow, with an additional efficiency improvement planned for next week.