Hi,
The daily free token allowance for the incentivized tier is 10 million. I was previously charged whenever my usage exceeded that limit. However, over the past few days, I have not been charged for overages, usage of 13 million or even 15 million tokens has still been counted as free.
Has the free token allowance been increased?
Thanks.
Hi and welcome back!
Sounds great from your perspective, and thanks for raising it anyway!
Can you confirm that you’re looking at the correct organization and project? Did you check this using the API or the dashboard on the platform?
There can sometimes be quite a delay before the data is displayed.
I believe I’m looking at the correct data. I checked both the raw JSON from the API and the web dashboard, and they are consistent.
As in the screenshot, before August 6, it appears that overuse was charged (shown in pink), but on later days, it was not. For example, on August 12, 15M was on the incentive tier, so this was clearly not just an overshoot.
Your screenshot combines the two large and small groups into one report: 15.3M input tokens, 128k output tokens - on unmentioned models (you can filter by model).
The cutoff for the first three days is visually at right about 10-11M, and then no more pink bar when going above that height sure would seem like “incentive tier” is catching everything, and you reconciled the $0 dollars billed as corroborating the experience.
Here’s the only complementary token help page even having gpt-5.6 mentioned, a page updated 11 hours ago (no version history).
The page still lists 10M tokens as the group free inclusion for Terra, Luna and other mini models in the larger pool. So to answer the topic’s title: there is no documented increase for everybody or anybody in particular.
In that page, there is some peculiarity in the language, that might make you think only input tokens are considered, which might just be a communication flaw.
For each new request, we check against your running total for that day. If a single request would make your daily total exceed the token quota, that entire request is billed at normal rates.
…and then there is a table of “tokens in request”…not “with tokens in response”. However, only input being counted cannot account for your 15M+ input alone on a day.
You’ve done your duty in reporting it to anyone listening. It’s not your duty to turn off complementary tokens and data sharing. Overbilling reports are also poorly responded to or remedied from clear bug reports on the forum.
Consider if you should persist ethically: you now are merely being compensated for all the training you are sending - instead of OpenAI still getting the excess as data sharing while you pay full price. Maybe they like what you are sending on API for their free perpetual future use?
Running a check with 10 million+ tokens led to the expected result of me getting billed after exhausting the incentivized tier.
Is it possible the costs showed up with a delay?
There are three days in the usage bar graph that would seem to be in excess of the complementary token budget, yet all the tokens are comped. It would be rather odd that any delay could instead re-classify a past week’s token totals.
It appears the bug is on the usage page but credits are deducted as expected.
That’s the question to @CJZ
@VeitB I checked again, and you were right. I was charged correctly, the bug is in the usage display.
The incentivized tier combines two pools, all types of tokens, and with models that are priced very differently, making the charges difficult to reconcile. I overlooked this in my earlier review.
So, the bug affects both the usage page and the usage API, not the billing.
Thank you all for your help. Unclear usage bothers me more than the actual charges.
Thank you for raising this!
I’ll forward our observations to the team to get this resolved.
I just did a quick comparison with the Usage API, and I can retrieve the expected data from there. That’s our workaround for the time being.
