I think I’ve hit a quota/account-state bug in the Codex app.
What I’m seeing:
- Standard Codex weekly usage limit: 0% remaining
- Standard Codex 5-hour usage limit: 100% remaining
- GPT-5.3-Codex-Spark weekly usage limit: 55% remaining
- GPT-5.3-Codex-Spark 5-hour usage limit: 100% remaining
However, the Codex app still shows:
“You’re out of Codex messages”
and blocks the composer, even when GPT-5.3-Codex-Spark is selected.
Why this looks wrong:
My understanding is that GPT-5.3-Codex-Spark is supposed to have its own separate usage bucket. But in the app, the exhausted standard Codex weekly limit seems to hard-lock the whole composer and prevents Spark from being used as well.
Expected behavior:
If GPT-5.3-Codex-Spark still has remaining weekly/5-hour quota, I should still be able to use it after switching the model selector to GPT-5.3-Codex-Spark.
Actual behavior:
The app appears to apply the exhausted standard Codex weekly limit globally and blocks Spark too.
Steps to reproduce:
- Use up the standard Codex weekly usage limit.
- Leave GPT-5.3-Codex-Spark weekly quota still available.
- Open the Codex app and select GPT-5.3-Codex-Spark in the composer.
- The app still shows “You’re out of Codex messages” and disables usage.
What I already checked:
- The usage dashboard clearly shows remaining GPT-5.3-Codex-Spark quota.
- The app UI still refuses input.
- I have screenshots of both the dashboard and the blocked composer state.
Environment:
- Plan: Pro
- Platform: [Mac app]
- Version: 26.313.41514
Question:
Is this a known bug in the Codex app quota gating logic, or is there some hidden rule where remaining GPT-5.3-Codex-Spark quota still cannot be used once the standard Codex weekly pool is exhausted?
I’m mainly posting to confirm whether others can reproduce this and whether OpenAI can fix the quota-state check in the app.