Weekly-limit percentage does not correspond to comparable token allocation across two Codex accounts
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 125k
- Forks
- 19.4k
- PR merge metrics
- PR metrics pending
Description
Summary
I use two separate Codex accounts, referred to here as Codex and Codex1. They are separate accounts, not different models or modes.
During the comparison workload, both primary accounts used:
- Model: gpt-5.6-sol
- Reasoning: max
- Mode: fast
Both accounts showed 100% weekly limit remaining at the start and 10% remaining when the comparison was made. However, Codex1 received materially less token usage than Codex.
The current Codex1 follow-up chat is using gpt-5.6-luna; that is a later configuration and is marked separately below.
Summary comparison
| Account | User chats | Subagent sessions | Models used* | Input | Cached | Uncached | Output | Thinking | Total |
|---|---|---|---|---|---|---|---|---|---|
| Codex | 8 | 14 | sol ×8, terra ×9, luna ×5 | 155,518,847 | 150,441,600 | 5,077,247 | 640,310 | 319,088 | 156,159,157 |
| Codex1 | 5 | 9 | sol ×4, luna ×4, terra ×6 | 75,250,655 | 73,071,616 | 2,179,039 | 302,457 | 151,879 | 75,553,112 |
| Combined | 13 | 23 | — | 230,769,502 | 223,513,216 | 7,256,286 | 942,767 | 470,967 | 231,712,269 |
*The model counts include spawned subagents. The primary comparison workload used gpt-5.6-sol with max reasoning on fast mode.
Codex1 received roughly half the total recorded token usage of Codex, despite both accounts moving from 100% to 10% of the weekly limit.
User-chat details
| Account | Start | Session | Type | Model | Input | Cached | Uncached | Output | Thinking | Total |
|---|---|---|---|---|---|---|---|---|---|---|
| Codex | 20:58 | 01a02f3c-8a9b | user | sol | 17,786,528 | 17,489,408 | 297,120 | 48,284 | 24,486 | 17,834,812 |
| Codex | 20:59 | 01a02f3d-9648 | user | sol | 8,362,927 | 8,167,424 | 195,503 | 35,186 | 18,461 | 8,398,113 |
| Codex | 20:59 | 01a02f3d-d42d | user | sol | 12,262,848 | 11,988,736 | 274,112 | 48,703 | 28,754 | 12,311,551 |
| Codex | 21:21 | 01a02f52-2804 | user | sol | 7,059,368 | 6,873,344 | 186,024 | 33,444 | 19,795 | 7,092,812 |
| Codex | 21:26 | 01a02f56-3b88 | user | sol | 9,530,692 | 9,325,824 | 204,868 | 43,187 | 22,064 | 9,573,879 |
| Codex | 21:30 | 01a02f5a-3e8a | user | sol | 10,896,438 | 10,557,184 | 339,254 | 61,980 | 32,635 | 10,958,418 |
| Codex | 21:42 | 01a02f65-8ab8 | user | sol | 23,749,331 | 23,267,712 | 481,619 | 77,577 | 37,578 | 23,826,908 |
| Codex | 21:47 | 01a02f69-e6f5 | user | sol | 15,120,838 | 14,744,320 | 376,518 | 48,423 | 25,605 | 15,169,261 |
| Codex1 | 20:59 | 01a02f3d-fda0 | user | sol | 15,823,269 | 15,561,728 | 261,541 | 37,734 | 19,394 | 15,861,003 |
| Codex1 | 21:00 | 01a02f3e-4c5c | user | sol | 15,021,264 | 14,727,936 | 293,328 | 47,837 | 27,778 | 15,069,101 |
| Codex1 | 21:18 | 01a02f4f-1da5 | user | sol | 13,019,698 | 12,776,704 | 242,994 | 47,946 | 30,124 | 13,067,644 |
| Codex1 | 21:25 | 01a02f55-59eb | user | sol | 8,139,868 | 7,918,848 | 221,020 | 37,962 | 19,406 | 8,177,830 |
| Codex1 | 22:29 | current follow-up | luna | 686,190 | 591,616 | 94,574 | 8,611 | 3,714 | 694,801 |
The remaining 23 sessions were spawned scout, test-writer, or implementer agents and are included in the summary totals.
Expected behavior
The weekly-limit percentage should either represent comparable token capacity across accounts, or the UI should expose account-specific quota, weighting, and remaining-token values.
Requested investigation
- Why did Codex and Codex1 consume materially different token amounts for the same displayed weekly-limit change?
- Does the weekly percentage account for model, account, service-tier, caching, or reasoning-weight differences?
- Can exact weekly quota and remaining-token values be exposed?
- Is this an expected account-level difference or a weekly-limit display/calculation issue?
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
The report names no files, tests, or entry points. Start by reproducing the two-account comparison and tracing how the weekly-limit percentage is calculated for model, caching, reasoning, and service-tier differences. Done means documenting whether the difference is expected or identifying a display/calculation issue, with account-specific quota information exposed if appropriate.
Written by the indexing model from the issue text.
Assessment
- Domain
- tooling
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100