←back to thread

507 points martinald | 2 comments | | HN request time: 0s | source
Show context
noodletheworld ◴[] No.45053394[source]
Huh.

I feel oddly skeptical about this article; I can't specifically argue the numbers, since I have no idea, but... there are some decent open source models; they're not state of the art, but if inference is this cheap then why aren't there multiple API providers offering models at dirt cheap prices?

The only cheap-ass providers I've seen only run tiny models. Where's my cheap deepseek-R1?

Surely if its this cheap, and we're talking massive margins according to this, I should be able to get a cheap / run my own 600B param model.

Am I missing something?

It seems that reality (ie. the absence of people actually doing things this cheap) is the biggest critic of this set of calculations.

replies(10): >>45053436 #>>45053533 #>>45053550 #>>45053564 #>>45053601 #>>45053730 #>>45053776 #>>45053962 #>>45055164 #>>45055610 #
jsnell ◴[] No.45053564[source]
> why aren't there multiple API providers offering models at dirt cheap prices?

There are. Basically every provider's R1 prices are cheaper than estimated by this article.

https://artificialanalysis.ai/models/deepseek-r1/providers

replies(1): >>45053818 #
ac29 ◴[] No.45053818[source]
The cheapest provider in your link charges 460x more for input tokens than the article estimates.
replies(1): >>45054086 #
1. dragonwriter ◴[] No.45054086[source]
> The cheapest provider in your link charges 460x more for input tokens than the article estimates.

The article estinates $0.003 per million input tokens, the cheapest on the list is $0.46 per million. The ratio is 120×, not 460×.

OTOH, all of the providers are far below the estimated $3.08 cost per million output tokens

replies(1): >>45054638 #
2. ruszki ◴[] No.45054638[source]
There are 7 providers on that page which have higher output token price than $3.08. There is even 1 which has higher input token price than that. So that "all" is not true either.