"You can now access GPT-4 models with 8K context via the existing OpenAI API" but maximum length is actually 2048

I joined the wait-list, and on Friday I had the pleasure to receive the email telling me that I’m invited to use GPT-4. However, the link in the email brings me here:

where I’m told again I have to join the wait-list. No biggie! I can just log into my OpenAI account and go to the playground. However, you may consider having the mail link point directly to the playground, rather than to the model description, to avoid any confusion.

Now, I have two questions:

  1. Which modes is GPT-4 available in? Only chat, right?
  2. The invitation email tells me that I can use GPT-4 with 8K context. However, the playground only allows me to crank the maximum length up to 2048 tokens. Why?
2 Likes

I joined the GPT4 waitlist and, on Friday or any other day, I did not have the pleasure to receive any email announcing that I could use GPT4 through the API. But it doesn’t make any difference between having access to the unusable OpenAI APIs and not having access at all, so.

IDK why they chose that max for the playground, you do get the correct max when calling the api directly.
Might be they just haven’t updated plaground.

I think it does support over 8k tokens. The slider is probably bugged or not updated, had it review a document that’s around 7k tokens and it was fine.

If you have issues using it you can check out this guide with common pitfalls and a javascript code example which is working.

Trying to use Autogen. Have just topped off my credit card payment, showing a positive balance. Is there some kind of high minimum balance to use gpt-4 with Autogen?