Repository navigation
Codex (gpt-5.5, Plus plan) — rate-limit cost per token jumped ~10-20x since June 16, draining the 5h budget in 2-3 prompts #28879
Description
Activity
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
- Codex 5-hour usage meter appears to consume much faster than comparable historical usage #28823
- Codex Desktop consumed 80% of 5-hour usage limit within 5 minutes #28498
- Codex quota was exhausted again during a short working session. #28727
- 5 Hour Usage drains afer just TWO messages because I recently purchased the credits #28065
Powered by Codex Action
Reacted by Ahmed Shahriar Sakib, Jaiby, Bruno , Eugene Lobach, abek, Ahmed Hindy, Jinhui Yuan and Samuel MalletSame here. I'm on the Pro plan and my quota suddenly vanishes in less than 5 hours now. Even with top-ups, it doesn't survive a single intense work session. This rate-limit spike completely breaks the entire workflow. Attaching my restriction message as well (referencing image_f0485a.jpg / my previous logs). Hopefully, this gets fixed ASAP.
Reacted by johnnyApplePRNG, Chris DiBussolo, i-fail, Will, SundayMoments, Ahmed Shahriar Sakib, kav-welg-4, Jaiby, AdAstra-LD, Z1xus and 29 more+1 here,
seeing a sudden and unusually large drain in my Codex 5h primary limit.
After a single simple conversational interaction, my 5h limit dropped from approximately 99% remaining to '67%'.. =(
This seems materially higher than expected for the type of interaction performed.
(Environment)
Plan: Plus
Surface: Codex Cli
OS: Linux
Model: gpt-5.5 high
Approximate time: 2026-06-18 22:20~30 KSTReacted by johnnyApplePRNG, Jaiby, Bruno , Mahi 14, Patrick Loboda, Ilya Novik, Martin Landa, JayPernet, Joe, AttieG and 6 moreReacted by João Rodrigo da Silva AfonsoSame here. Even more than that - after a few or a dozen prompts throughout the day, without any large documents (at most a dozen or so kilobytes) and without using MCP, my profile shows, for example, 22-26 MILLION (!) tokens used. And I have reasoning set to Low or, at most, Medium. This is crazy! At these prices, if I had to pay for tokens, I’d be paying for the 5.4 model (since I don’t even use 5.5) at $15 per million - that’s over $300 a day! I think they’re being calculated incorrectly. This can’t be accurate information about token usage.
Reacted by Bruno , saket, Patrick Loboda, sidprax, Ilya Novik, Martin Landa, Jinhui Yuan, signadou and DanielsaanReacted by Peter Jun Koh, Ilya Novik, signadou and DanielsaanI had to switch to Claude; now Codex seems like a more aggressive Claude 🤣. Now I feel like Claude doesn't use any resources. Codex is unusable today.
Reacted by Jihwan Oh, LI Long, Lukasz D, Thomas Trenty, Delvin Bacho, Ilya Novik, abek, Seb, devoren, Morpheus and 6 moreReacted by Bruno , Ilya Novik, abek, ThinkWithPbody and DanielsaanI'm experiencing the same behavior on a Pro 20x account.
Since approximately June 18, my 5-hour limit has started draining at a rate that is completely inconsistent with historical usage. The weekly quota is decreasing proportionally as well.
Before this change, I was never able to exhaust my 5-hour allowance, even during heavy Codex usage sessions involving large codebases and long-running tasks. Since yesterday, the same workflows are consuming the entire 5-hour budget in roughly one hour, which is much closer to the behavior I previously observed on a Plus plan.
Nothing has changed on my side:
- Same account and subscription tier (Pro 20x).
- Same workflows and usage patterns.
- No significant increase in prompt size or task complexity.
- No changes to models or configuration.
The timing closely matches the reports in this issue and other recent quota-related reports. From the user perspective, it appears either:
- The effective weighting of usage against the 5-hour budget changed around June 16–18, or
- The quota accounting system is currently overestimating usage.
Given that multiple Plus and Pro users are reporting the same sudden regression over the same time window, this looks more like a platform-wide accounting or rate-limit issue than a change in individual usage patterns.
Would be helpful to know whether any quota calculation changes were deployed around June 16–18 and whether the current behavior is expected.
Reacted by Peter Jun Koh, Sascha Greuel, milos, Jaiby, William Mitchell, Bruno , Thura Htet Aung, saket, Yegor Shulga, Ali Dogus Temizsoy and 7 moreSame pattern here, from a paid Pro/Codex/Hermes setup. I posted the detailed trace in #28823, but the short version is:
2026-06-19 19:06 AEST backend endpoint: primary 5-hour used: 52% secondary weekly used: 59% weekly reset: 2026-06-25 08:46 AEST plan_type: prolite Spark separate bucket: 16% weekly used / 0% 5-hour usedThis is less than one day into the weekly window and after only small recent interaction counts. OpenAI’s public Codex pricing says Pro is “5x or 20x higher rate limits than Plus,” while Plus GPT-5.5 local messages are published as 15–80 / 5h. The practical behavior now feels materially inconsistent with that representation unless there is an undisclosed reweighting/multiplier or mis-accounting bug.
The product also pushed an “upgrade because you are nearing limits” style message, which is not an adequate answer for an already-paid plan when the issue is unexplained quota drain.
What users need is a per-turn/per-feature ledger: model, client/surface, bucket charged, input/cache/output/reasoning tokens, tool/MCP/image/retry/startup/compaction/background charges, and any multipliers.
Detailed trace: #28823 (comment)
Reacted by Peter Jun Koh, Bruno , Jehadur Rahman (Emran), Pete W, Renzo , Heller, Martin Landa, Seb, jkooooo99, Jinhui Yuan and 2 more+1
Reacted by Bruno , gazeciarz, Jehadur Rahman (Emran), jkooooo99, Jinhui Yuan and DanielsaanSame here on a ChatGPT Pro / Pro 5x account.
I primarily use Codex through the CLI. Over the last ~2 days, the 5-hour usage window has started draining much faster than normal, despite my workflow being basically the same as before.
Reacted by Leonardo Mazzaferro and Martin Thorsen RanangSame issue here. Saw this start to happen when the 2x offer ended, and the remaining was not half.
This happened to me today as well. Curiously, right after I started working, I noticed my tokens were being used up very quickly relative to the five-hour usage window. I'm currently having the Pro plan.
Does anyone know what might be going on?
+1 to this, I'm a power user in Mac. I'm seeing this since June 1 drastically increase in rate limitsvl consumption. I'm on 100 dollar plan and I see for 5 hour session I'm usually finished the limit in 1.5 hour majority of the times
188 remaining items
i had to move on Claude, same pattern, high cost for same prompt that I send everyday. WT* are happening?
i had to move on Claude, same pattern, high cost for same prompt that I send everyday. WT* are happening?
The squeeze?
Windows 11
Codex Plus Plan
GPT 5.5 LowI just wrote the word "Hello" and left the PC. Codex has used up 20% of the 5 hour limit and 5% of the weekly limit. For saying hello to him.
No MCPs are enabled. There are no agents and other things. I'm using Antigravity now,
GOODBYE CodexSame here I am using plus plan and it is evaporating during discussion session after 10 questions
Is it just me, or does Codex still feel like it's using a lot more of my quota compared to a month ago. Yes, it's better than a two weeks ago when we had a massive problem, but it still drains my quota faster nowadays.
Reacted by Adam B and Renzo100%.
I have a workflow I've been using for months.
Previously I could chat with codex for ~1 hour before being rate limited.
Now I'm lucky to get 6 questions before using my 5hr quota, using the exact same prompts and skills.
I still have the sessions in codex from before this started showing this huge disparity.
Currently browsing OpenRouter for alternatives, this is ridiculous.
today the 5 hour limit is gone and since just a little bit ago limits seem to be ok for me again (im on 99% now for quite some time while yesterday some simple prompts ate a lot of pro lite....
The 5-hour limit removal is currently touted as temporary. With the ChatGPT5.6 rollout, Codex App -> ChatGPT App migration, and announcement of 5.4 retirement... it's been an absolute roller-coaster of quota resets, random performance, and your-guess-is-as-good-as-anyone regarding results weekend.
This is blatant "move fast and break things" strategy (with a bit of Claude vs Codex PR thrown into the mix). Don't bother testing, or releasing a coherent - or even stable - product; just shite it out the door any time it suits Marketing - and have subscribers PAY to do the testing and work through the pain points. //slow-clap
Same Weekly Codex tokens gone in two days with a pro account.
I'm burning through my weekly limit (that used to last a week) in a day (50% of my weekly usage in ~2 hours so far today). I'm downgrading myself to Terra for now and hoping this gets fixed, otherwise I'll need to supplement with Claude.
My 'turns per week' has dropped from ~300 previously to ~70.Summary
Starting ~June 16, my ChatGPT Plus Codex budget drains in 2–3 prompts on
gpt-5.5, where the same model/plan/app previously gave 20+ prompts. My Codex session logs (token_count/rate_limitsevents) show the limit-% consumed per token has increased roughly 10–20×, with no change on my side — prompts are actually smaller now and reasoning tokens are near zero.
(...)Same Problem!
Win 10
GPT Work
4 persistent tasksHow can I maeassure the metrics form my account like you: Prompt, Input tokens, Reasoning, tokens Primary limit %?
Thanks!For those that missed it: https://x.com/thsottiaux/status/2082317452755751098
Cross-linking this same-model/same-plan pre/post comparison to #41220, a meta tracker for abnormal Codex usage/quota depletion and usage-accounting inconsistencies. This report is important to the cluster because the claimed ~10–20x change occurs on GPT-5.5 with smaller prompts and near-zero reasoning in the affected period, so the broader symptom cannot be dismissed as merely GPT-5.6 Sol/Ultra being expensive. The meta tracker does not presume a single root cause.
Summary
Starting ~June 16, my ChatGPT Plus Codex budget drains in 2–3 prompts on
gpt-5.5, where the same model/plan/app previously gave 20+ prompts. My Codex session logs (token_count/rate_limitsevents) show the limit-% consumed per token has increased roughly 10–20×, with no change on my side — prompts are actually smaller now and reasoning tokens are near zero.Environment
gpt-5.5,service_tier = "default"plus(confirmed inrate_limits.plan_type)Evidence — same plan, same model, same app
June 12 (working — 20+ prompts/day):
→ ~57K-token prompt with 3.6K reasoning ≈ 1% of the 5h budget
June 18 (broken — 2–3 prompts):
→ ~20K-token prompt with zero reasoning ≈ 10–27% of the 5h budget
Per-token limit cost changed ~10–20×. Smaller prompts, near-zero reasoning, yet each consumes 10–20× more of the 5-hour budget than a 57K-token xhigh prompt did on June 12.
What I ruled out (verified from logs)
model_reasoning_effortwas alreadyhigh/xhighon the working days (June 12–15). xhigh gave 30–40+ prompts on June 3, 12, 13, 15.Corroborating timing
OpenAI status page shows an active incident ongoing ~2 days (since ~June 16), which matches exactly when my usable prompt count dropped from ~9 sessions/day to ~4.
Ask
Was the Plus-tier
gpt-5.5rate-limit budget reduced or the per-token weighting changed around June 16? If intentional, what's the new budget? If not, can it be investigated?Session log references
rollout-2026-06-12T09-07-36-019eba71-5c57-70d0-a8fa-aa999731e9ff.jsonlrollout-2026-06-18T10-22-14-019ed99b-ea9d-7fa1-a86b-dca4e55a8e3e.jsonltoken_countevents withrate_limitspayload (fields:primary.used_percent,secondary.used_percent,plan_type,resets_at).