The Usage page used to show you exactly how many tokens you were using. While it wouldn’t show how many tokens per individual request, it’d group requests together based every hour, showing you the total input tokens, total output tokens, and a total sum of both each hour of each day.
Now, I don’t see any way of seeing how many tokens you’re using per request; just how much money you’re spending each day. You can’t see how many tokens are being used anymore.
And you also cannot see how many requests are being made either; well, you can – but only for image, audio, and assistant models, and fine-tuning.
Whilst I can view what models I’m using and how much each costs me each day, I am no longer able to access the useful information of how many requests are being made each hour of each day, how many input and output tokens, respectively, are being used each hour of each day, and how many total tokens are being used each hour of each day.
I don’t wish to complain, as I love OpenAI’s services, but it seems quite backwards to remove featuress that were highly useful. I don’t understand why these features of the Usage page were removed, and I’d love to have theme back.
Just chiming in that it would have be really nice to have those features back. I’m currently having to add code to count input and output tokens and make database calls just for debugging. It’s not a huge deal, but it’s an extra step compared to just checking the Usage page
More likely: OpenAI already looked at it, saw how it would look for assistants to show up as dozens or hundreds of calls, and also the API barrage they would get from Stripe in polling to get back what was consumed, and implemented the change deployed the same day as their devday release of assistants.
Personally, I do think it’s a fairly big deal. I think that it’s important to address that transparency from OpenAI regarding our spending is crucial. The previous Usage page offered valuable insights, and they removed helpful features from the page, which means managing our spending is more difficult. It’s another reason not to go with OpenAI, as other services provide in-depth, detailed activity pages that help manage spending and deciding on what models to use.
I love the services that OpenAI provides, but it is important to be able to see how many input tokens, output tokens, and total tokens each request uses and to be able to see detailed information for each day and each hour of each day like we could before. I don’t understand why they had to remove these features, and I hope they can be re-added.
I don’t disagree. I only said it wasn’t a huge deal, because those token counts are available in the response body. Just more work to parse them and save them somewhere
The “Credit Grants” bar on the usage page is incomprehensible to me and thus useless. I would prefer to see how much money is left in my account before my API requests start returning “too many requests”.
The info I want is in there somewhere; I found it yesterday. But today I went looking for it again and couldn’t remember how to find it. It should be right on the usage pages.