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

[mihneaptu]

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_limits events) 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

• Codex desktop app, Windows 11
• Model: gpt-5.5, service_tier = "default"
• Plan: plus (confirmed in rate_limits.plan_type)
• Config unchanged across the regression window

Evidence — same plan, same model, same app

June 12 (working — 20+ prompts/day):

Prompt	Input tokens	Reasoning tokens	Primary limit %
06:08	22,334	488	2%
06:09	52,161	1,130	3%
06:51	57,112	3,645	9% → 11%
06:54	79,961	5,125	15% → 16%

→ ~57K-token prompt with 3.6K reasoning ≈ 1% of the 5h budget

June 18 (broken — 2–3 prompts):

Prompt	Input tokens	Reasoning tokens	Primary limit %
07:22:27	18,910	152	18%
07:22:30	20,480	0	45%
07:22:34	23,469	48	57%
07:22:43	23,631	0	68%
07:22:50	23,839	0	77%

→ ~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)

• Not reasoning effort: reasoning tokens are 0–152 today vs 3,645–5,125 on the good days.
• Not prompt/context bloat: input tokens dropped from 57K–80K to 18K–23K.
• Not my config: model_reasoning_effort was already high/xhigh on the working days (June 12–15). xhigh gave 30–40+ prompts on June 3, 12, 13, 15.
• Not the weekly cap: secondary window is only at 12%. The 5-hour window is the one draining.

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-5.5 rate-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

• Good day: rollout-2026-06-12T09-07-36-019eba71-5c57-70d0-a8fa-aa999731e9ff.jsonl
• Bad day: rollout-2026-06-18T10-22-14-019ed99b-ea9d-7fa1-a86b-dca4e55a8e3e.jsonl
• Rate-limit data is in token_count events with rate_limits payload (fields: primary.used_percent, secondary.used_percent, plan_type, resets_at).

[github-actions]

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

[eva763057345-lab]

Same 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.

[MDGChamomile]

+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 KST

[StrumykTomira]

Same 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.

[Aesthermortis]

I 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.

[andr3sdr]

I'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:

1. The effective weighting of usage against the 5-hour budget changed around June 16–18, or
2. 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.

[Truck0ff]

Same 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 used

This 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)

[avneetranjan]

+1

[caioarotolo]

Same 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.