›_ Codex Issues
Codex Issues / #18355

Codex says limit reached even though usage still shows 2% remaining

Resolved 💬 2 comments Opened Apr 17, 2026 by rafaself Closed Apr 17, 2026
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What issue are you seeing?

Codex shows that I still have 2% of usage remaining, but when I try to send a new prompt it says I have already reached the maximum limit.

<img width="613" height="892" alt="Image" src="https://github.com/user-attachments/assets/ae18518e-512e-40e1-96bf-cb1f86ced1dc" />

---

<img width="1276" height="240" alt="Image" src="https://github.com/user-attachments/assets/3d39bffb-3547-4b9c-a897-cde55924a82d" />

What steps can reproduce the bug?

  1. Open Codex with usage showing 2% remaining.
  2. Try to send a new prompt.
  3. Codex blocks the request and says the maximum limit has already been reached.

What is the expected behavior?

If usage still shows 2% remaining, Codex should allow new prompts until the limit is actually reached.

Additional information

_No response_

View original on GitHub ↗

2 Comments

github-actions[bot] contributor · 4 months ago

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

  • #17345
  • #18083
  • #18234
  • #16592
  • #17537

Powered by Codex Action

emmanuelmailukyalo · 3 months ago

"limit reached" message is a common issue with Codex due to the distinction between token count and usage quota (5-hour or weekly limits).

As of April 2026, the system heavily factors in "reasoning time" and request complexity rather than just the raw number of characters.
Here are the primary reasons why this happens:

  1. High-Cost "Reasoning" vs. Token BalanceThe 50%+ figure often refers to your total context tokens, but the limit reached message refers to your 5-hour usage window.Heavy Requests:

Complex coding tasks, large file refactoring, or high-intelligence models (like GPT-5.4) burn through your 5-hour quota much faster than they consume raw context tokens.Fast Burn Rates:
A few complex queries can consume 100% of the 5-hour allowance even if you have used only 10% of your weekly token limit.Memory Tool Overhead:
If you are using the memory tool, it can consume 100% of your 5-hour window via background summarization.

  1. Time-Based Limits (5-Hour Window)Codex enforces a strict limit every 5 hours.The "5-Hour" Wall: If you make many requests in a short time, you will hit this cap regardless of how many tokens are left in your total weekly quota.Not Just Token-Based: It is not a simple "max tokens per day" system. It is a "workload per time" system.

i have 53% tokens and it is saying i have reached quota