You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
What terminal emulator and version are you using (if applicable)?
WSL
What issue are you seeing?
Reposting this since you failed to read and comprehend the last one. Understand the following:
There's a huge discrepancy between regular limit usage and credit usage.
This has DRASTICALLY changed with the rollout of gpt-5.3-codex
I've been working on a refactoring plan and started using Codex 5.3 yesterday. Most of the work is iterating over the same requests, where I ask Codex for architectural improvements, performance improvements, simplifying things and so on.
With regular limits it cost maybe 1-2% of my weekly limit (on Plus).
Now that my weekly limit is reached, Codex is gobbling credits like crazy. I have maybe 4 of those same requests eating more than 100 credits. To put this in perspective, working on the same codebase I'm used to 100 credits being roughly one day's work. At this rate, I'm not even sure 1000 credits would take me through the day.
Refactoring tasks typically incur significant token usage, especially if you are using "high" or "xhigh" reasoning effort.
I don't have access to your token usage stats, but the credit consumption that you're reporting sounds in line with what I'd expect.
If you find that the credit pricing doesn't align with the value that you're receiving from the tool, another option is to switch from a Plus to a Pro subscription. This will increase your usage limits by about 10x.
Read the issue description again. What exactly is unclear?
I'm repeatedly running the same tasks to tighten the spec.
Regular limits are used up slowly
Credits are used up at an insane rate
This does not match the behavior from previous versions
Lastly, as a paying customer I expect reports of an issue to be read, understood and taken seriously. Not glossed over and dismissed. Please do better.
I have seen the same issue. I think it's essentially that the 5.3 version burns through the "budget" much faster. I noticed that at my normal coding rate with the 5.2, when I would burn through the budget given by the standard Plus subscription, the additional credits purchased for $40, had a similar burn rate to the standard Plus subscription amount. In the sense of "It takes 6 hours of using 5.2 to use my Plus budget, and if I buy more credits for $40, it takes maybe another 6 hours." But after the 5.3 update, it feels like "I burn through my Plus subscription budget of credits/tokens in 6 hours, and for $40, I burn through the extra credits in 3 hours."
I chalked it up to OpenAI subsidizing the Plus budget to scale it up, but not bothering to change it for the extra codex credits you can buy, and so the get burned through much faster.
Activity
etraut-openai commentedon Feb 8, 2026
Refactoring tasks typically incur significant token usage, especially if you are using "high" or "xhigh" reasoning effort.
I don't have access to your token usage stats, but the credit consumption that you're reporting sounds in line with what I'd expect.
If you find that the credit pricing doesn't align with the value that you're receiving from the tool, another option is to switch from a Plus to a Pro subscription. This will increase your usage limits by about 10x.
For more details about Codex pricing, refer to this documentation.
oldandangry commentedon Feb 8, 2026
Read the issue description again. What exactly is unclear?
Lastly, as a paying customer I expect reports of an issue to be read, understood and taken seriously. Not glossed over and dismissed. Please do better.
wandadars commentedon Feb 11, 2026
I have seen the same issue. I think it's essentially that the 5.3 version burns through the "budget" much faster. I noticed that at my normal coding rate with the 5.2, when I would burn through the budget given by the standard Plus subscription, the additional credits purchased for $40, had a similar burn rate to the standard Plus subscription amount. In the sense of "It takes 6 hours of using 5.2 to use my Plus budget, and if I buy more credits for $40, it takes maybe another 6 hours." But after the 5.3 update, it feels like "I burn through my Plus subscription budget of credits/tokens in 6 hours, and for $40, I burn through the extra credits in 3 hours."
I chalked it up to OpenAI subsidizing the Plus budget to scale it up, but not bothering to change it for the extra codex credits you can buy, and so the get burned through much faster.
kdawgwilk commentedon Feb 18, 2026
Is this more related to the temporary 2x limits on the plans right now?
MalcolmNsibande commentedon Apr 27, 2026
I'm experiencing the same issue. Any update on this?