Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

OpenAI dropped the price of o3 by 80%

twitter.com

Post reply on HN
Print view View original

491–500 of 518 posts

Re: OpenAI dropped the price of o3 by 80%

#491

Post by ethbr1 »

fsmv wrote: post #425
prmoustache wrote: post #409

Earlier quoted context omitted.

The fact they don't sell ads doesn't mean they are not in the business of selling users data to third parties. Also Netflix wasn't initially selling ads and there you have after increasing the price of their plans drastically in the last few years the ad supported subscription is probably the #1 plans because most people aren't willing to shed 15 to 25usd/€ every month to watch content that is already littered with a…

If you sell ads you're actually incentivised not to sell data because then your competitors would be able to reach your users without paying you

You're incentivized not to sell targeting data, but you're very incentivized to collect and maintain as much of it as you can, and then offer access using it as a service.

So, at the end of your day, company X has an overdetailed profile of you, rather than each advertiser. (And also, at least in the US, can repackage and sell that data into various products if it chooses)

Re: OpenAI dropped the price of o3 by 80%

#492

Post by otterley »

numlocked wrote: post #435
ddtaylor wrote: post #314

Earlier quoted context omitted.

I also am using OpenRouter because OpenAI isn't a great fit for me. I also stopped using OpenAI because they expire your API credits even if you don't use them. Yeah, it's only $10, but I'm not spending another dime with them.

Hi - I'm the COO of OpenRouter. In practice we don't expire the credits, but have to reserve the right to, or else we have a uncapped liability literally forever. Can't operate that way :) Everyone who issues credits on a platform has to have some way of expiring them. It's not a profit center for us, or part of our P&L; just a protection we have to have.

Out of curiosity, what makes you different from a retailer or restaurant that has the same problem?

Re: OpenAI dropped the price of o3 by 80%

#493

Post by otterley »

scudsworth wrote: post #147
simonw wrote: post #139

Earlier quoted context omitted.

Sounds like you want their Zero Data Retention plan: https://platform.openai.com/docs/guides/your-data#zero-data-... (It's "contact us" pricing, so I have no idea how much that would set you back. I'm guessing it's not cheap.)

it doesn't seem like this would supercede a court order

The court order doesn't require OpenAI to modify their software. ZDR is implemented through a separate API with separate endpoints that never retained data in the first place.

Re: OpenAI dropped the price of o3 by 80%

#495

Post by godelski »

sgarland wrote: post #420
jazzyjackson wrote: post #276

Earlier quoted context omitted.

(having no insider info:) Because it can be used as a primary key ID across aggregated marketing databases including your voting history / party affiliation, income levels, personality and risk profiles etc etc etc. If a company wants to, and your data hygiene hasn't been tip top, your phone number is a pointer to a ton of intimate if not confidential data. Twitter was fined $150 million for asking for phone numbers…

Tangential: please do not use a phone number as a PK. Aside from the nightmare of normalizing them, there is zero guarantee that someone will keep the same number.

Also fun fact, people mistype and mistranscribe data. Some people even... put down fake phone numbers because they don't want to be tracked!

I would think in a world where we constantly get spam calls and texts that people would understand that a phone number is not a good PKI. I mean we literally don't answer calls from unknown numbers because of this. How is it that we can only look at these things in one direction but not the other?

Re: OpenAI dropped the price of o3 by 80%

#496

Post by godelski »

azinman2 wrote: post #332
jazzyjackson wrote: post #276

Earlier quoted context omitted.

(having no insider info:) Because it can be used as a primary key ID across aggregated marketing databases including your voting history / party affiliation, income levels, personality and risk profiles etc etc etc. If a company wants to, and your data hygiene hasn't been tip top, your phone number is a pointer to a ton of intimate if not confidential data. Twitter was fined $150 million for asking for phone numbers…

OpenAI doesn’t (currently) sell ads. I really cannot see a world where they’re wanting to sell ads to their API users only? It’s not like you need a phone number to use ChatGPT. To me the obvious example is fraud/abuse protection.

  > To me the obvious example is fraud/abuse protection.
Phones are notorious for spam...

Seriously. How can the most prolific means of spam be used to prevent fraud and abuse? (Okay, maybe email is a little more prolific?) Like have you never received a spam call or text? Obviously fraudsters and abusers know how to exploit those systems... it can't be more obvious...

Re: OpenAI dropped the price of o3 by 80%

#497

Post by beering »

alternatex wrote: post #352
mrcwinn wrote: post #286

Only at HN can the reaction to an 80% price drop be a wall of criticism.

"80% price drop" is just a title. The wall of criticism is for the fine-print.

What in the fine print are we criticising? Most of the negative comments make no reference to any fine print on their website.

Re: OpenAI dropped the price of o3 by 80%

#498

Post by beering »

hadlock wrote: post #74
xmprt wrote: post #29

Earlier quoted context omitted.

How are LLMs cached? Every prompt would be different so it's not clear how that would work. Unless you're talking about caching the model weights...

I've asked it a question not in it's dataset three different ways and I see the same three sentences in the response, word for word, which could imply it's caching the core answer. I hadn't previously seen this behavior before this last week.

Isn’t the simpler explanation that if you ask the same question, there’s a chance you would get the same answer?

In this case you didn’t even get the same answer, you only happened to have one sentence in the answer match.

Re: OpenAI dropped the price of o3 by 80%

#499

Post by azinman2 »

godelski wrote: post #496
azinman2 wrote: post #332

Earlier quoted context omitted.

OpenAI doesn’t (currently) sell ads. I really cannot see a world where they’re wanting to sell ads to their API users only? It’s not like you need a phone number to use ChatGPT. To me the obvious example is fraud/abuse protection.

> To me the obvious example is fraud/abuse protection. Phones are notorious for spam... Seriously. How can the most prolific means of spam be used to prevent fraud and abuse? (Okay, maybe email is a little more prolific?) Like have you never received a spam call or text? Obviously fraudsters and abusers know how to exploit those systems... it can't be more obvious...

It costs money to get a phone number. It’s about friction, not elimination.

What would you do instead?

Re: OpenAI dropped the price of o3 by 80%

#500

Post by fastball »

IanCal wrote: post #392
fastball wrote: post #349

Earlier quoted context omitted.

Anecdotal, but about a week ago I noticed a sharp drop in o3 performance. For many tasks I will compare Gemini 2.5 Pro with o3, running the same prompt in both. Generally for my personal use o3 and G2.5P have been neck-and neck over the last months, with responses I have been very happy with. However starting from a week ago, the o3 responses became noticeably worse, with G2.5P staying about the same (in terms of wha…

Are you sure you're using the same models? G2.5P updated almost exactly a week ago.

G2.5P might've updated, but that's not the model I noticed a difference. o3 seemed noticeably dumber in isolation, not just compared to G2.5P.

But yes, perhaps the answer is that about a week ago I started asking subconsciously harder questions, and G2.5P handled them better because it had just been improved, while o3 had not so it seemed worse. Or perhaps G2.5P has always had more capacity than o3, and I wasn't asking hard enough questions to notice a difference before.

Post reply on HN