Aware that you need to spend $1 minimum on 3.5-turbo, however I’ve spent over that (although only 12 hours ago), and wondering how long people have had to wait before getting access to GPT-4?
The $1 in billing doesn’t matter anymore.
You’ll get GPT-4 access when the resources exist to provide it. New developers are being granted access every week.
Do we have to do anything else to get access to GPT-4, or will we automatically grant permissions to users using GPT-3.5. Thank you.
Thank for your answer, I got it. I’ll continue to wait then.
I’ve created a brand new account and added the credit card. Then I have the GPT-4 access from the start.
Take a look at my rate limits.
There is one change to the registration process. I was required to deposit a minimum of $5 while adding the credit card. It’s different from the temporary $5 hold before. This is the actual charge.
The rate limit is even more generous than my old account. My old account with payment record has 10,000 TPM/200 RPM limit.
If you’re in a hurry, I suggest you create a new account.
Is this true? Has anyone tried this?
If it is true, that means despite making current developers wait due to a lack of resources, they are prioritizing resources for new registrants. That’s quite dishonest.
- Apparently the user above tried it.
- There’s nothing “dishonest” about it.
This works - I just added a CC for billing and bought $5 credit (minimum order) and GPT-4 immediately showed up in the Playground.
The fact your “new” user got access before others with more milage, maybe related to openai’s understanding that gpt4 api users would most probably stop the plus subscription.
@nomuhyuna What page is that that shows all your models and rate limits?
I am using the playground to see if I have access to gpt-4 (which I don’t)
Apparently the rate limit page (which I just found out about from this thread): OpenAI Platform
Interestingly, I have the same “base level” 40,000/60 TPM/RPM for GPT-4, but I have:
90,000/3500 for GPT3.5 (4k models) and
180,000/3500 for GPT3.5-16k
Also, my page does not show TPM/RPM/RPD, but only TPM/RPM. I can’t imagine 3500 RPM is correct for GPT3.5.
After 48-hours the is no longer a requests-per-day limit.
Here are my chat rate limits for reference,
The 3500 RPM is correct for
That said, I think their rate limits are a little off. It’s very hard to hit the RPM limit before the TPM limit even if you’re not doing any batching of requests (with the exception of embeddings, I hit the RPM limit all the time with that endpoint).
You’ve got the gpt-4-32k, and the gpt-4 caps don’t change? Bummer.
How to buy credits? I cant find any option to do that.
Good to know. I’m trying to rack up $1 of charges using gpt3.5 and then wait until the next billing cycle to use gpt4. The tool I’m waiting to try wants gpt-4. I guess it doesn’t help.
I think it was the “Payment methods” button to add a credit card:
It was definitely on this screen, but it was a little wonky - the screen I thought I needed to go to kept telling me to go somewhere else. Sorry I don’t remember the specific steps.
You can’t use the same phone number for a new account unfortunately.
Older account with untouched API.
I must commend on the new “upgrade” call to action and easily-discoverable “start payment plan”.
A prepayment option or requirement isn’t immediately obvious when then taken to the first “input credit card” screen. Perhaps someone willing to go further can document or screenshot options after adding the card.
“Payment methods” has text “You have not started a payment plan yet.”, whereas before (or for existing payers) that is where a post-pay credit card was entered or updated.
(Now see if adding $1000 prepay (if given such an option on a new account) actually gives you a higher hard limit out of the gate…)
OpenAI seems to have twice added additional GPT-4 users in bulk, the first with the early July announcement, and second mid-August based on July billings.