I’m eager to gain access to the GPT-4 8k API, but I’m a bit confused about the process. According to this article, a $1 deposit is required. However, when I navigate to my OpenAI account, I can’t find an option to deposit funds.
Can someone guide me on where and how to make this deposit? Any help would be greatly appreciated!
Thank you in advance.
SOLUTION:
I just closed my plan and “Billing”->“Overview” and opened it again. When I reopened it, I was able to make a contribution to the API. At least 5 $.
For a new account, you would need to “start payment plan” on your account’s billing options. The minimum credit purchase is actually $5.
For an older account, future consideration of onboarding is those who qualify by have actually been billed over $1 to their credit card on file from non-trial usage. billing history. It doesn’t immediately unlock access.
I have a ChatGPT Plus subscription, but my API account does not have access to GPT-4. Are they different accounts? My billing history shows zero since Mar 20, 2023. How do I “start a payment plan” when I am already on billing, and have a ChatGPT Plus subscription?
The cost of using API is by your actual use of data.
The billing system is completely separate from a ChatGPT subscription.
If you’ve already put in a payment card method for API, but simply not used it enough in the past to get a bill, you have a post-paid account. You would not want the prepaid “buy credits” plan in its current state of bugs and inconvenience. Just generate >$1 usage causing billing from usage to put yourself in the “not casual trial moocher” category for future consideration.
3.5-turbo
Model Input Output
4K context $0.0015 / 1K tokens $0.002 / 1K tokens
So I’d need to use 1000 tokens to pay $2, about 750 words they say.
Thank you _j
Update: I’ve done three translations, each consuming 1000 tokens, and have $0.0016 used so far. It’s pretty efficient, so I won’t be hitting any PAYG threshold soon.
No need to incur billing? Inaccurate, or at least poor advice to deny its relevance. More access to GPT-4 was granted in a second wave after the announcement and after he wrote that - again only to those with account billings.
Hi, that is indeed out of date advice, there was an update recently where users with more than $0.50 in API usage credit (pre pay account) were granted GPT-4 access, from that one can reasonably make the assumption that at least a bill of $0.50 is required to enable GPT-4 access on a post pay account.
With that all being said, there is no official word on the exact mechanism by which those who pay for API usage on their due billing date (traditional API users) gain access to to GPT-4. So do not make any commercial decisions reliant on that.
OpenAI would benefit from allowing to simply pay $0.5 instead of having users waste compute on nonsense generation, and users would benefit from not wasting own time.
I’d hope for their sake they have priced these APIs so their trajectory is to make money.
But my main point here is:
GPT-4 access during development phase is seriously overrated (and a costly exercise!).
By the time you get to Production I guarantee you’d have spent enough to get into the next wave of GPT-4 access
But once in Production, you might want to consider the additional cost and slowness of GPT-4 response to be a significant downside which might persuade you to stick with GPT 3.5. (but of course it really depends on what you are using it for and if the model differences matter).
I’m somewhat baffled by the number of posts asking how to get to $1 of API usage to enable GPT-4 access. First, if you want to set money on fire quickly, I think the you can just call a “text-davinci” model which is like half the cost of gpt-4.
But also, there is no prompt experimenting, function call refinement, testing, development to do that won’t consume a few dozen gpt-3.5 calls? Whatever you are planning to do with gpt-4, start doing it with gpt-3.5 and start evaluating. Once you have access to both models, you can determine what, if any, part of your pipeline can be fed to 3.5 at 1/20 the cost of gpt-4.
OpenAI completions deprecation announcement: “chat is 97% of our use”.
So a ChatCompletion script is more likely at hand for the $0.11 usage crowd. An even higher hurdle for useful usage: fine-tune a model.
Cancelling post-pay and paying for a credit means you’ve yet to use AI services, yet the money is not in your bank any more, and you have enrolled in limiting prepay system with possible bugs to encounter.
Guess the average amount of abandoned credit per account in a $5 minimum purchase scheme in a few years…
So you lose a buck, on minimum-impact call like I describe (and actually get the AI service you use). You’ve maintained your post-pay status by not removing cards and re-enrolling, and are also in the crowd “I spend money”.
@AlexKr I think you should reply to it as a post and self-accept, if that’s possible, since I usually don’t read the questions at all and have completely missed your solution, which indeed works. I’ll paste it for convenience:
I just closed my plan and “Billing”->“Overview” and opened it again. When I reopened it, I was able to make a contribution to the API. At least 5 $.
This is ingenious, it seems AI also has many faults. But in fact this really is a global problem, because very often with payment systems there are various problems, not long ago I found this source https://casinosanalyzer.com/online-casinos/cashlib that helps me find the payment method that suits me, this makes me very happy, because in the past I couldn’t play because I couldn’t cash in the casino account.
