Pro limits now feel dramatically worse and are getting exhausted in ~2 days

Resolved 💬 12 comments Opened Mar 20, 2026 by tully-8888 Closed Mar 20, 2026
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

Over the past few months, the usage limits for Codex/ChatGPT on Pro feel like they have become dramatically more restrictive.

What used to be enough for sustained, serious work is now being exhausted extremely quickly. In my case, usage that previously would not have been fully consumed even with very heavy use is now reaching 0% remaining within around 2 days.

This is creating a bad experience for Pro users, especially those using Codex for real development workflows rather than occasional prompts.

What feels wrong

  • The available quota appears much lower than it used to be.
  • The rate at which usage gets consumed feels disproportionate to the actual amount of work being done.
  • For a paid Pro plan, the current limits make longer coding sessions hard to rely on.
  • The product feels much less usable for serious iterative development than it did a few months ago.

Why this is a problem

Codex works best in extended back-and-forth workflows such as:

  • exploring codebases
  • iterating on implementations
  • fixing regressions
  • refining prompts and plans
  • continuing multi-step sessions over time

When the quota disappears this quickly, it breaks the workflow and makes the tool unreliable for professional use.

Request

Please:

  • clarify whether limits for Pro have changed recently
  • provide more transparency around how usage is calculated
  • reconsider the current limits for Pro users
  • offer a higher-cap option for users who rely on Codex heavily for development

At the moment, the current limits feel far too restrictive for a paid tier, and noticeably worse than the experience from a few months ago.

If this is intentional, clearer communication would help. If it is a regression or policy change, I hope it can be revisited.

View original on GitHub ↗

12 Comments

github-actions[bot] contributor · 4 months ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

  • #15039
  • #15094
  • #15166
  • #15073
  • #15140

Powered by Codex Action

daniel-leonte · 4 months ago

I encountered the same problem, quota was running out faster than usual on Pro accounts.

passc0de · 4 months ago

+1

I am encountering the same issue; my Pro account is hitting limits quickly.

swaroop-24 · 4 months ago

+1

I'm experiencing the same issue, my Pro account quota ran out in just 1.5 days

4erdenko · 4 months ago

Absolutely the same problem on the Pro account.

easwar03 · 4 months ago

+1

Enormously inconveniencing to the point of discontinuing my subscription. The quota which used to last 5d now lasts under 2d.

SureshKumar-Rahul · 4 months ago

The recent change in quota usage has become a significant bottleneck. My current allocation is depleting in under 48 hours—less than half the time it used to last. This drop in utility is making it difficult to justify my ongoing subscription

GhostPixel404 · 4 months ago

I am encountering the same issue; my Pro account is hitting limits quickly.

FilipDlouhy · 4 months ago

I have two subscriptions of pro and whey are both spent few weeks ago i would last on one now i need two if this continues i will cancel my subscriptions. Be at least honest with your customers openAI.

VannaGamma · 4 months ago

Again, as in the other multiple threads: also affected.

zuuxuux · 3 months ago

Definitely encouhntering this too

dogancanbaris · 3 months ago

This is at an unbelievable rate at the moment. Fast off, the regular context window. I used to go for a full week with codex 5.4 xhigh nonstop in like 5 projects and couldn't hit limits. Right now, Using 3 and in 5 hours down 15% using gpt 5.4 high. I know 30% difference in cost between those too but this is definitely a problem with the harness or cache.

No subagents, no memory nothing. Just running is vanilla, and usage limits are draining extremely fast. Needs to be definitely looked at