side not, wtf is wrong with this forum, i cant put dollar signs in my text?, comes out weird
Hey, @Zoinks, welcome to the forum!
The input and output tokens are calculated separately at diff prices…
Price
Input:
$1.250 / 1M tokensCached input:
$0.125 / 1M tokensOutput:
$10.000 / 1M tokens
Were you using reasoning?
the prices you posted are not for MINI, i am using mini
no, i just simply call the chatgpt api and gave it the gpt5-mini model as shown in the screenshot, u can see its using mini model for most of the amount spent, and those numbers dont align with official pricing
now i wonder how much they have been overcharging me in total using other models
Ah, sorry about that.
Are you logging the tokens when you make the calls? That might help explain it more.
Just using gpt5-mini, I believe reasoning is on by default (set to medium) which might be where your extra tokens are coming from?
I am not logging tokens, , the number of tokens is fine, 2.4m, its the pricing, isnt it just 2.4m x 2.00 dollar? for output? i thought it was just that straight forward
as far as i know I am not using reasoning of any kind, just straight up calling api with that model
i was using regular 5 before mini and it seems to be same price for mini so not sure what whats going on, the screenshot should have all the output and input tokens numbers plus the pricing charge for diff models, and thats where it dont make sense, unless i am mistaken and its not straight up token numbers X output price for determining output cost
The power of AI means you don’t have to make mistakes in your math, you let the AI do that for you.
| Model | Input Cost | Cached Input Cost | Output Cost |
|---|---|---|---|
| gpt-5-2025-08-07 | $0.005 | – | $0.198 |
| gpt-5-mini-2025-08-07 | $0.579 | $0.002 | $8.892 |
| gpt-5-nano-2025-08-07 | $0 | – | $0.013 |
Using your prices-per-million and the line items you shared, we re-derived tokens, limited by dollar figure accuracy (and verified in Python):
| Model | Input | Cached input | Output | Model total |
|---|---|---|---|---|
| gpt-5 | 4,000 | — | 19,800 | 23,800 |
| gpt-5-mini | 2,316,000 | 80,000 | 4,446,000 | 6,842,000 |
| gpt-5-nano | 0 | — | 32,500 | 32,500 |
| Total | 6,898,300 |
Breakdown:
The UI’s “Total tokens” appears to be input-only
Why UI input (2,408,624) is slightly higher than billed inputs (2,400,000)
Two effects can explain the +8,624 delta, and both are consistent with your notes:
Free-tier input tokens are counted in the “Total tokens” widget but not in the cost line items until you overflow.
Rounding of displayed $ amounts (shown to $0.001 precision) can alone account for a difference of this size.
Example sensitivities of a $0.001 change → token error:
Given these bounds, 8,624 is well within expected rounding noise if dollars are rounded before display.
No evidence of output tokens in that UI number
Inferred tokens used (from billing):
UI “Total tokens”: 2,408,624
Does it disagree?
Not once we interpret the widget as input-only tokens. The remaining +8,624 is plausibly explained by free input tokens that didn’t show in costs and/or rounding of the displayed $ amounts. Under either explanation, the small delta is reconcilable; there’s no sign of a billing inaccuracy in the data you provided.
If you can export the underlying (unrounded) dollar amounts or a per-direction token breakdown from the UI, we can pin down exactly how much of the +8,624 is free-tier vs rounding.
looks like the mistake was completely mine
if i hover over the total tokens,
it lterally tells me the input token at 2.409m
and output 4.497M
and with those numbers the charge seems to be correct or at least makes more sense than the intiial 2.4m total token, I read it as total token for output/input combined. that label seems misleading, or at least i think so
No worries. It happens to the best and worst of us at one point or another. Haha.
I hope you stick around. We’ve got a great community growing, and a wealth of knowledge.
However, this report IS valid: “Total Tokens” is incorrect.
Just checked my usage.
The “Total Tokens” is only summing the selected period’s input tokens. There is no way to get it to actually include or compute the output tokens over the date range. In the bar graph appearing below that, the output tokens can only be viewed by-day by hovering, never totaled.
The UI needs a bugfix.
There is a way, by clicking “chat completions” bellow in the usage page (or in this link):
Then, choose the date range and composition as you wish:
There will be a total for the whole period, above the chart.
(although the terminology “total tokens” remains wrong at the usage page)
Over at my place, I have a welcome mat inviting all in for safe passage.
Go into the bathroom, open the medicine cabinet, and read the post-it note on the back of the mirror that actually says you are in danger…
The information presented right up front is false with no way to rectify it.
