Codex Plus: 5-Hour Usage Limit Returns
CodexOpenAI reinstated the rolling 5-hour usage limit on Codex and ChatGPT Work for ChatGPT Plus subscribers, effective August 25, 2026, after roughly six weeks without it. The cap runs alongside the existing weekly quota rather than replacing it, so Plus users who burn through a session quickly can be locked out until the window resets even with weekly capacity left. OpenAI's Codex/ChatGPT Work technical lead, Thibault Sottiaux, said the change is meant to smooth server load and protect newer users from accidentally exhausting their entire weekly allowance in one sitting. The $100 and $200 Pro plans remain exempt from the 5-hour cap for the time being.
Key Takeaways
- OpenAI reinstated the rolling 5-hour usage cap on Codex and ChatGPT Work for ChatGPT Plus subscribers, effective August 25, 2026, after about six weeks without it.
- The 5-hour window runs alongside the existing weekly quota, not instead of it — Plus users can hit the five-hour ceiling even with weekly capacity left.
- Thibault Sottiaux, OpenAI's technical lead for Codex and ChatGPT Work, announced the reversal on X on August 24, citing server-load smoothing as the main driver.
- The cap was originally lifted in mid-July 2026 around the GPT-5.6 Sol launch, giving Plus users an uncapped stretch before this reinstatement.
- Pro subscribers ($100/$200 plans) are exempt from the five-hour limit for at least the next few months, widening the gap between Plus and Pro for long agentic coding sessions.
- The announcement drew swift pushback, including a front-page Hacker News thread and multiple OpenAI Developer Community posts from users who say it severely restricts long Codex sessions.
Sources & Mentions
5 external resources covering this update
OpenAI restores 5-hour Codex and Work limits for ChatGPT Plus users
Hacker News
OpenAI restores 5-hour Codex and Work limits for ChatGPT Plus users
9to5Mac
ChatGPT Plus is getting throttled again, and it's apparently for your own good
Digital Trends
Bad news: ChatGPT Plus users have annoying limits once again
Android Authority
Bring back the 5h limit for Plus accounts across ChatGPT Work and Codex
OpenAI Developer Community
Why the Limit Is Back
OpenAI removed the rolling five-hour usage cap on Codex and ChatGPT Work in mid-July 2026, around the launch of GPT-5.6 Sol, leaving ChatGPT Plus subscribers bound only by a weekly quota for about six weeks. That uncapped stretch ends August 25, 2026: OpenAI is reinstating the five-hour limit for anyone on the $20-per-month ChatGPT Plus plan.
How the Cap Works
Once a Codex or ChatGPT Work session starts, a rolling five-hour window opens, and everything Codex does inside it — prompts, file reads, tool and terminal output, cloud tasks, and model responses — counts against the allowance. The five-hour cap sits alongside the pre-existing weekly quota rather than replacing it, so Plus subscribers now face two ceilings at once: burn through the five-hour window and work pauses until it resets, even if the weekly total still has room left.
Why OpenAI Says It Made the Change
Thibault Sottiaux, OpenAI's technical lead for Codex and ChatGPT Work, announced the reversal on X on August 24, framing it as a load-smoothing measure that lets OpenAI keep weekly Plus allowances relatively generous. He also pointed to newer, more casual users who were unknowingly burning through an entire week's quota in a single long session and then finding themselves locked out for days — the five-hour window is designed to catch that pattern earlier and spread usage out more evenly.
Who Is Affected
The restored limit applies only to the Plus tier ($20/month). OpenAI says the $100 Pro and $200 Pro plans will not have the five-hour cap turned on for at least the next few months, preserving their larger, less-restricted allowances. Developers running long, continuous Codex sessions — for example, working through a Friday night into the weekend — may find Plus is no longer sufficient for that workflow and will need to budget around the new window or move to Pro.
Reaction
The change generated a swift and largely negative reaction, drawing a front-page Hacker News discussion and multiple threads on the OpenAI Developer Community forum from users who say the reinstated cap "severely restricts" their ability to run longer agentic coding sessions in Codex.