I’ve so far never had any luck reaching the Help Center, so I figured ask here in case anyone knows: Is there any place where you can see if and how much remaining credit/balance you have?
I’m asking, I’ve been using the API the past month, got my bill from the March and saw that the amount I’ve used has been paid by ‘balance’. I didn’t know I had balance, I can’t find any information on said balance. If I had known that I wouldn’t get a charge on my credit card and instead used this ‘mysterious balance’, I’d have made a lot more use of the API.
So does anyone know where this balance come from and where I can see how much is left?
3 Likes
Same here, it’s probably the worst costumer support I’ve ever experienced
I’m glad we can help each other on this forum, the page your looking for is under account → usage, here’s a direct link:
Thanks. It doesn’t tell me anything though regarding credits, only usage. Last month, it counted my normal usage. (with the remark: Total usage so far this month.
This includes all usage, even usage that may be offset by credits. Accordingly, your monthly bill may be lower than the value shown.) I looked at it almost daily and it counted normally, no weird things happening.
To my knowledge I do not have credit. (as I have a pay as you go account, of which it says: You’ll be billed at the end of each calendar month for usage during that month.) And there is nowhere to see if there is credit and if so, how much there is.
On the receipt it says Total: $…
Applied balance -$…
which leaves me with an invoice of $0.00
I just think it’s weird that if there is any credit or balance, it’s showing up nowhere. I’ve looked at every page in my billing overview and there’s just no information on it. 
1 Like
It is important that we monitor our budget!
I do this with 4 methods:
1.Account menu then usage limit
That’s where I set my max budget and alert
2.Account menu then usage to often check the usage
3.Sometimes, when using the GPT-3.5 turbo and GPT-4 APIs intensively, I parse the json response object to display the tokens.
4.BEFORE running intensive tasks, I estimate the words and convert them to an estimated number of tokens (1 word is about 0.75 token). Then I go to the pricing menu and estimate the cost.
For an important project comes up, I recommend automating this functionality in a classical function.
I hope this helps.
Great post!
OpenAI has a website for counting tokens:
If you want to tokenize text within your own code, you can check out our tiktoken package for Python. For JavaScript, there’s the gpt-3-encoder package for node.js.
This post is 80 tokens
Agreed, cost estimation is hard because you do not have good usage statistics before you launch your service.
The way we handle access to computational resources at our institution is by giving everyone a set amount of credits/tokens when they log in the first time. They can then apply for a grant when they run out of credits. This has worked decently for our ~11.000 students and employees