Summary
It seems that the report is multiplying by x1000 the actual token usage and estimated cost:
➜ npx cxusage
| date |
events |
input_tokens |
output_tokens |
total_tokens |
est_cost |
| 2025-09-13 |
1488 |
1,659,482,361 |
9,856,510 |
1,669,338,871 |
$1086.46 |
| 2025-09-14 |
1918 |
6,963,991,542 |
27,910,470 |
6,991,902,012 |
$4492.05 |
| 2025-09-18 |
3305 |
2,163,789,550 |
22,078,387 |
2,185,867,937 |
$1462.76 |
| 2025-09-19 |
4482 |
2,824,862,435 |
31,865,027 |
2,856,727,462 |
$1924.86 |
I'm on a Plus plan so it doesn't make sense. I noticed that when I burn around 300k (cached) in the codex status bar, I see an increase of about 300,000,000 total tokens.
As for the price, if you divide by 10, it's around 5.5$ / week (2 days for me) which makes sense for a 20$ plan.
cxusage version
No response
Python version
No response
Steps to reproduce
Relevant output/logs
Summary
It seems that the report is multiplying by x1000 the actual token usage and estimated cost:
➜ npx cxusage
I'm on a Plus plan so it doesn't make sense. I noticed that when I burn around 300k (cached) in the codex status bar, I see an increase of about 300,000,000 total tokens.
As for the price, if you divide by 10, it's around 5.5$ / week (2 days for me) which makes sense for a 20$ plan.
cxusage version
No response
Python version
No response
Steps to reproduce
Run `npx cxusage`Relevant output/logs