Hello, no API access with gpt-4 model, although there is a history of successful payments long before July 7, 2023.
When trying to use the gpt-4 model, the official openai python client returns an error -
“openai.error.InvalidRequestError: The model:
gpt-4 does not exist”
although everything works fine with the gpt-3.5-turbo model
I have the same issue. It says that we should have had access by July 7 (I sent a receipt and invoice showing a successful payment in May), but it seems that is not actually correct for many people.
End of month is for people who have not had a successful payment:
Today all existing API developers with a history of successful payments can access the GPT-4 API with 8K context. We plan to open up access to new developers by the end of this month, and then start raising rate-limits after that depending on compute availability.
Successful w/ my payment but showing this using aider: Aider v0.11.1-dev
API key does not support gpt-4-32k, falling back to gpt-3.5-turbo-16k
I am unable to utilize gpt-4 using aider.
Welcome to the OpenAI developer forum,
What is “Aider v0.11.1-dev”? it does not sound like something OpenAI makes, also the 32k model is not publicly released.
same issue after one successful payment cycle completed and also played for chatgpt plus ,still showing no access to gpt-4 ,for another account
Same here, I have paid for API usage and understand that it’s opened up at the beginning of this month. Still I don’t have access (not when trying to use the model through the API nor the playground).
Does anyone have a clue as to what can be the cause?
Subscribing to ChatGPT Plus ($20/month) has no effect on the API, its payment considerations (“as an active developer”), or the API feature set enabled. ChatGPT is the web-based chatbot, and plus unlocks features like GPT-4 and plugins just within ChatGPT.
The OpenAI API is immediately available to anyone that submits a payment method for their monthly pay-per-use bills of the API service within their account. It doesn’t take a subscription or pre-authorization beyond a valid credit card in the completely-separate billing system.
The language in the GPT-4 API announcement a month ago was that they would start opening up GPT-4 to new (read “more”) API developers at the end of July, not specifically all developers. We can see that the new developers that have been since onboarded is minimal, likely based simply on the management of limited compute resources (and the opposing demand from $240/year ChatGPT users that have no wait list).
TL:DR; You don’t have access to GPT-4 because of OpenAI’s choice, not OpenAI’s error.
The OpenAI API is immediately available to anyone that submits a payment method for their monthly pay-per-use bills of the API service within their account.
That’s incorrect. I and many other have had valid payment methods set up for months, have used more than $1, and yet still don’t have access as of August 7.
What you say about OpenAI choosing to grant or not grant access is correct; it is indeed unlikely to be an “error”. But I think many people are frustrated about OpenAI’s low-quality communication.
At the moment, OpenAI can still afford to treat their (wannabe) customers like farm animals. But as the field expands, and other (and possibly better) options become available (e.g. Claude, the Meta one, etc.), consumers will be quicker to jump ship. The only reason I’m still patiently waiting is because the main tool I use is currently dependent on OpenAI’s API, but as soon as the makers of that tool manage to offer other, non-OpenAI solutions, I’ll be the first to sign up.
I’m sorry if I was not clear. I am not talking about GPT-4 access in that particular numbered statement, but rather contrasting API to ChatGPT – and that there is no “subscription” to API.
“Subscription” language is often discovered in these long topics, that show that people are conflating ChatGPT payments with API payments and entitlements.
You can check your billing history here, see that you were charged money for non-free use before the May-June cutoff of the first determination.
There is no evidence that making billings and payments after the initial cutoff date would or will be again considered for GPT-4 eligibility (but it can’t hurt to show you are a serious user).
Yes, I know you were talking about the API, and that we are both talking about “access” to, rather than a subscription to, that API. I do have a $20/month subscription to ChatGPT, but I am trying to get access to the API, specifically for GPT-4 (as GPT-3.5, which I do have access to, is not of much use to me).
And I also agree that “there is no evidence that making billings and payments after the initial cutoff date” will guarantee access, but OpenAI’s help pages still say: “On July 6, 2023 , we gave access to the [GPT-4 API] (8k) to all API users who have made a successful payment of $1 or more,” which at least suggests that making such a payment is a precondition to getting access.
And that was kind of my larger point: why are they being so opaque? Why can’t they just say what the current status is, and what’s likely to happen in the next few weeks/months? Which gets to my other “larger point”, namely that they are being opaque because they can be, because they have no serious competition at this point, but that tide may well turn quickly, leaving a large number of disgruntled (and former) OpenAPI users in its wake.
Yes, frustrating and opaque are pretty good words. I was checking to see if my account still in trial grant got turned on: no.
“We plan to” can certainly mean that those plans went amiss.
The numbers of new developers that did get GPT-4 access and took advantage could be more than allows further expansion.
They could be faced with no ability to cut off acceptance of new “plus” subscriptions without serious bad publicity, with successful subscription numbers there also wanting the GPT-4 compute resources (which should have been opt-in instead of default).
Competition from others building their own AI datacenters and consuming a limited supply of advanced GPUs, and even Microsoft keeping their Azure ML cloud just for themselves and enterprise users, could impact plans – even precipitating a rollback of AI output quality and quantity to manage existing resources.
Opaque seems to be the name of the game. A tweet “we didn’t make GPT-4 dumber”? I disagree, with proof. Any statement that could be made probably goes to a team of corporate lawyers first.
@logankilpatrick Same here, have reached the $1 expense on API but still getting
openai.error.InvalidRequestError: The model
gpt-4 does not exist or you do not have access to it. Learn more: How can I access GPT-4? | OpenAI Help Center.