The Codex burned through my entire limit. #44673

Activity

1. NickitaChe

   added

   appIssues related to the Codex desktop appIssues related to the Codex desktop app

   on Sep 10, 2026
2. github-actions

   added

   bugSomething isn't workingSomething isn't working

   rate-limitsIssues related to rate limits, quotas, and token usage reportingIssues related to rate limits, quotas, and token usage reporting

   on Sep 10, 2026
3. github-actions commented on Sep 10, 2026

   

   github-actionsBot

   on Sep 10, 2026 – with GitHub Actions

   Contributor

   More actions

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

   • Codex app usage astronomically higher than CLI #44459

   Powered by Codex Action
4. digitdash commented on Sep 10, 2026

   

   digitdash

   on Sep 10, 2026

   More actions

   This is not really helpful unless you at least provide what model and effort you were using along with the prompt(s) used to consume the majority of your usage limit.

   If you were using Astra / Ultra, this can definitely explain it. Ultra gives the agent the option of spawning subworkers as needed, all on max effort. So if the orchestration agent decided to use, say, 5 sub workers for a task, you'd literally be paying around 5x in usage for the same amount of time compared to using a single agent. This in conjunction with using the flagship model can eat up even the highest $200 plan within about half a day. This is from my own experience at least.
5. NickitaChe commented on Sep 10, 2026

   

   NickitaChe

   on Sep 10, 2026

   Author

   More actions

   @digitdash I use the Astra Medium tier for very simple tasks (fixing code, executing CMD commands). Previously, on the Sol High tier, these tasks consumed 30–40% of the weekly quota; however, following the Astra release, consumption has become excessive—now, a single task lasting 10–25 seconds consumes at least 1%.
6. digitdash commented on Sep 10, 2026

   

   digitdash

   on Sep 10, 2026

   More actions

   @digitdash I use the Astra Medium tier for very simple tasks (fixing code, executing CMD commands). Previously, on the Sol High tier, these tasks consumed 30–40% of the weekly quota; however, following the Astra release, consumption has become excessive—now, a single task lasting 10–25 seconds consumes at least 1%.

   Hmm, yeah that's very strange. I will say that Astra for me seems to use nearly double the same as previous. Medium seems to be the way to go with Astra though, which you can see on codexradar.

   Have you tried using Sol ever since the update? I'm curious to know whether usage remains as before or if its also affected for you.
7. NickitaChe commented on Sep 11, 2026

   

   NickitaChe

   on Sep 11, 2026

   Author

   More actions

   Here’s a bit of proof: it gained a percent in two minutes. Let me try to keep going with a high SOL
8. NickitaChe commented on Sep 11, 2026

   

   NickitaChe

   on Sep 11, 2026

   Last edited by NickitaChe

   Author

   More actions

   It’s the same problem with Sol: the same "2 minutes for 1 percent" issue... That only gives me enough for 4 hours of work out of my weekly limit... Before, I could easily put in 40 hours, and there was still enough left for personal needs.

   
9. taobaibais commented on Sep 12, 2026

   

   taobaibais

   on Sep 12, 2026 – with ChatGPT Codex Connector

   More actions

   Same kind of failure here on ChatGPT Plus today (September 12, 2026 around 21:00 CEST).

   My Codex 5-hour allowance was showing 100% remaining. I resumed only a few very small tasks/follow-ups, then shortly afterward it was suddenly 0% remaining / fully exhausted.

   I was not using GPT-6 Astra. The visible work was far too small to explain burning an entire fresh 5-hour window. This looks like another current quota/accounting/reconciliation failure rather than normal usage.
10. mizumico commented on Sep 27, 2026

    

    mizumico

    on Sep 27, 2026

    More actions

    Your "1% per 2 minutes" measurement matches the published billing rates almost exactly: Astra is metered at ~2.5× Sol per token, and post-launch consumption was further distorted by the Sept 9–10 counter incident (#44199). So part of the gap is real pricing, part is that incident's broken accounting — worth re-measuring now that the counters have stabilized before concluding Sol itself got 10× worse.

    On the parts you can control:

    1. Fast mode off. The blue lightning toggle bills ~2–2.5× for a 1.5× speedup — worst trade on the menu.
    2. Astra Low for routine tasks. Independent measurements show Astra Low already beats Sol High on quality; Medium/High on "fix this code, run this command" tasks is where weekly quotas go to die.
    3. Context management. [features.context_management] with experimental_mode = true in ~/.codex/config.toml cuts long-task consumption noticeably.

    If after all that your honest workload needs more than the plan gives, the math usually favors pay-per-token over upgrading: the CLI's [model_providers] config (wire_api = "responses", CLI ≥ 0.122) runs the same models with no weekly ceiling, and at documented API rates your ~4-hours-of-Astra week would cost single-digit dollars instead of a plan jump.