Have you heard of anyone getting it in August/official word from an employee?
Seems like no one i know has gotten it rolled out
Yes, I’ve heard from several people who have gained access and one who posted a reply from support outlining the need to balance new onboarding with usage.
1 Like
If you are unsure:
For ii = 1:100:
{Run API call with a bunch of tokens}
%% wait 10 mins to see what the cost is, check your account management page, and repeat if needed
1 Like
Strange, been waiting since may and we have credit that expires in may of next year as the grant was for one year.
Still nothing 
1 Like
With all due respect, was that not considered when openai.com announced on July 6th that all new developers would have access by the end of July?
My developer dashboard indicates that I have a paying account, and I have a valid credit card saved as my payment method. I’ve not been billed for a charge >= $1, so I understand that I don’t qualify through that method. I’ve been on the waiting list for months, and figured you were rolling out API access slowly because you needed to scale up computing power, which makes sense.
But then why announce that there would be general availability by the end of July, then say access is rolling out slowly because of load issues? Was there some setback? I think people are very frustrated with the lack of clarity and transparency in the very generic explanations. I think most would prefer an accurate explanation of when they can expect general availability, even if the timeline has been pushed back since the initial July 6 announcement.
4 Likes
The best way to view this is that now, developers no longer need to apply for GPT-4 API access, it will be granted automatically as soon as possible, I wish everyone could get access all at the same time, but if you give everyone access and the service performance drops, then developers will complain, and at the same time, if you restrict access to keep performance acceptable you again get complaints. It is being rolled and and you will get access.
2 Likes
i have already consumed 1.45$ but the gpt4 api has not been activated.
Do I have to wait for the payment to be made at the end of August?
I need access to api gpt 4!
2 Likes
From what I understand, you need to wait until you have been successfully billed once by OpenAI. So at the end of this current billing cycle, if you’re using GPT 3.5 and creating costs, you will be given access. I think.
The $1 billing is no longer a thing. It doesn’t matter how much you’ve had in billing. That was a threshold for an earlier cohort.
They are currently adding access for more developers to gpt-4. There is nothing anyone needs to (or can) do to speed up their access to the gpt-4 model.
All that is left to do is wait.
2 Likes
Are you sure about that? I’ve had a running conversation with OpenAI Support about this through the middle of July and they indicated that:
- you must exhaust your trial credit (even if you find the GPT-3.5 API to be unusable for purpose, like the original poster and myself);
- you must accrue at least a $0.50 charge above and beyond your trial credit so that the charge posts to your billing method;
- the billing cycle completes at the end of the month;
- invoices may not be issued for up to two weeks after the end of the month.
The exact quote from one of their replies is as follows:
Once your usage reaches at least $0.50, our system will trigger a charge, granting you access to GPT-4 8k and its advanced capabilities. Please note that your usage will be invoiced at the end of each calendar month. Invoices are typically issued two weeks after the end of the billing cycle. Finishing the free trial and reaching the minimum usage requirement will ensure you have seamless access to all the benefits of GPT-4 8k engine.
That particular reply was on July 12th, nearly a week after the announcement.
(In my case, I finished burning through the trial credit by generating text embeddings that I needed for some vector DBs and then burned a couple more dollars beyond on that, just to be safe. That was all done by July 22nd. So, now, I guess I “just wait”.)
I believe some people who had >$1 spending are complaining as well of not having GPT4 API access yet.
I’m using gpt3.5 model and the quality is awful its understanding of requests is by far inferior. I think I’ll just create that $1 charge and forget about this , moving forward to test openAI’s competitors. how many tokens does one need to exhaust to trigger that $1 charge anyway?
1 Like
The gpt-3.5-turbo-16k model is probably going to be the most “efficient” just-give-me-gpt4-already token burner at $0.003 / 1K tokens, according to the pricing page. So, $1 / $0.003 = 334 * 1K tokens = 333,334 input tokens, if I did the math correctly. (Not counting the price of generated tokens, which can be controlled with the output token limit parameter.) I was much less efficient with the text embeddings, since they cost $0.0001 / 1K tokens, but I had an actual need to fill with those while I was waiting for GPT-4 API access and I actually did need to process millions of tokens.
I think that there is $5 credit, so actually I need to spend 1,500,000 tokens. yes the payment is also on the output tokens. maybe one of them is more expensive than the other. I don’t even care, let it be $10 bill this is such a nonsense to invest time in
Will there be an announcement once GPT4 is available? Should we subscribe somewhere to be notified?
I am not sure it is practical to make an announcement for each user being onboarded, and an announcement to say that a batch has been processed would also not give any new information.
Best thing is to check regularly in the Playground for the additional models.
Are longest waiters prioritized, or is selection random? You also say “for developers” - is “developer” anyone who has an OpenAI account?
This Help Center page also needs updating - I believed I had to spend $1 so I spammed a bunch of tokens. (Yes, if read carefully, it doesn’t say “spend $1”, but that’s the implication.)
I am not familiar with the onboarding access priorities, I would imagine it’s a combination of waitlist position and API past usage.
1 Like
dear manager maybe you can edit youre lie to some True ?
becoz you saund like GPT 2.1 now 
Well, it likely doesn’t take waitlist time into account. I’ve been waiting since March after signing up and paying for usage on the API. I’m about to give up on waiting at this point.
I finally got access within the past hour and they did send an email, as promised when I first joined the waitlist in early May. Not sure if that is what you meant by “will there be an announcement”, but it seems they are notifying individuals when access is granted to them, so checking the playground might not be necessary.
(Also, I was granted access within a day of being invoiced for the past month’s usage, so it seems that the response I shared from OpenAI Support earlier in this thread was nearly true… at least in my case. I’m still a bit displeased by the shift from “join on our waitlist to get access” to “spend money with us to get access” and the fact that this was poorly communicated.)
3 Likes