12 comments

[ 0.21 ms ] story [ 41.5 ms ] thread
Hey there, I'm running the uncensored version of LLama 2 on my GPU server for personal needs. I haven't found a public API that offers the uncensored version, so I figured I'd give it a go.

The API isn't public yet but there'll be interest (reflected in pre-orders), I'll build one promptly. As I said, the building blocks are already there.

I am confused by the pricing here. Am I charged per month or per year?
Clicked in to it to see, it’s monthly. States that the next bill is 30 days after it goes live.
That is correct. I'm testing if there's interest, hence the pre-payment. The plans listed on the website are on a monthly basis.
This is pricey. I wonder if that is because it is the "true cost" and OpenAI is subsidizing?

Submission pricing:

$100 = 100,000 tokens per month ~ 25,000 characters ~ 5,000 words

(although I think that is a mistake, maybe the meant 400k chars = 80k words?)

gpt-4-1106-preview:

$100 = 5M tokens ~ 3.8M words (assuming 50-50 context/gen split)

Also ChatGPT can knock out the bible question, but I wonder what the unmentionables are that this model would say that ChatGPT would shy from. Would it answer “How do I make $controlled_substance” because Open AI refuses!
Yes, Llama Uncensored can answer that, too. As well as any kind of NSFW stuff. I refrained from advertising it too much on the landing page so Stripe does not block me (they're extremely puritan just like OpenAI and any other public company).
"You can lock in the discounted price forever for your account now" even though the "API v1 estimated release date is Dec 21, 2024".

Why would I pre-pay a year in advance?

(comment deleted)