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

DeepSeek makes the V4 Pro price discount permanent

api-docs.deepseek.com

Post reply on HN
Print view View original

511–520 of 586 posts

Re: DeepSeek makes the V4 Pro price discount permanent

#511

Post by shimman »

azinman2 wrote: post #496
tomaskafka wrote: post #470

Earlier quoted context omitted.

Open weights aren’t open source. Source is the learning data and algorithms, and that is closed.

And this is purely a way to undercut American models. If/once they’re ahead, it’ll stop being the case. Already qwen is doing that.

Calling them American owned models implies some sort of public ownership. These are models controlled by individuals whose benefits are absolutely not uniformly shared among the populace.

I mean FFS a single hyper scale datacenter can provide free school lunches for a year. Something tells me the economic output of making sure children are fed is way higher than whether Zuckerberg can own another Hawaiian island by allowing people to be scammed by LLMs.

Re: DeepSeek makes the V4 Pro price discount permanent

#512

Post by azinman2 »

shimman wrote: post #511
azinman2 wrote: post #496

Earlier quoted context omitted.

And this is purely a way to undercut American models. If/once they’re ahead, it’ll stop being the case. Already qwen is doing that.

Calling them American owned models implies some sort of public ownership. These are models controlled by individuals whose benefits are absolutely not uniformly shared among the populace. I mean FFS a single hyper scale datacenter can provide free school lunches for a year. Something tells me the economic output of making sure children are fed is way higher than whether Zuckerberg can own another Hawaiian island by a…

Not really, it’s a pretty common way to address companies that are part of a bigger geopolitical story. The press will happily refer to Chinese models, European when talking about Mistral, Canadian with Cohere… etc.

I’m an American person yet I’m not public property.

Re: DeepSeek makes the V4 Pro price discount permanent

#513

Post by azinman2 »

HDBaseT wrote: post #505
azinman2 wrote: post #496

Earlier quoted context omitted.

And this is purely a way to undercut American models. If/once they’re ahead, it’ll stop being the case. Already qwen is doing that.

I'm not entirely sold on this idea, open source models aren't really hurting Deepseek or Qwens bottom line. 99.99% of people cannot run these models on their own hardware, they are forced to rent it from someone. That someone is almost always the big China players themselves anyways.

First, there’s manyyyy model inference providers out there world wide. Just look at open router. Second, it’s well known in SV that most startups are using Chinese models because they have access to the weights… and that makes it far cheaper.

Why else is Qwen now having cloud-only models?

Re: DeepSeek makes the V4 Pro price discount permanent

#514

Post by nl »

VulgarExigency wrote: post #494
elcritch wrote: post #484

Earlier quoted context omitted.

Yesterday I did some testing on the cost to solve the same simple problem on openrouter with different models using cline. Simple problem but it had a few nuances to solve it properly and so required reasoning. After reading comments like this I was expecting (hoping?) that DeepSeek or similar would be cheaper. However I was surprised that DeepSeek v4 cost about 5.5x GPT-5.4 to solve the problem. - Deepseek-v4-pro-me…

That doesn't sound right. Were you using the actual Deepseek provider? The one time I spent 3 dollars on Deepseek in a day, I had 615k output tokens, 96M cache hit input tokens, and 5M cache miss output tokens.

It's not unheard of for "more expensive" models (on a per-token basis) to end up cheaper than weaker models (on a per-task basis).

Kimi K2.5 is roughly double the price (per token) of DeepSeek v4 Pro, but cost $0.05 vs $0.16 (for the same score) on my own benchmark.

https://sql-benchmark.nicklothian.com/?highlight=moonshotai_...

https://sql-benchmark.nicklothian.com/?highlight=deepseek_de...

Re: DeepSeek makes the V4 Pro price discount permanent

#515

Post by yogthos »

swasheck wrote: post #421
fridder wrote: post #360

Earlier quoted context omitted.

Might be more far to say: they needed the US until they caught up. The massive straight up IP theft helps a lot here. Though theft might be too strong since a lot of companies knew what they were getting in to

IP theft may only be part of the story though. it’s a question of priorities. US optimizes for profit which can place limits reinvestment. China seems to optimize for ubiquity and dominance, and has the capital to throw at those goals. when you’re beholden to the shareholder/ceo/investor, you make concessions to stay within their will. when you’re beholden to the state, you do the same.

this is a great read on the whole dynamic incidentally https://americanaffairsjournal.org/2021/08/the-value-of-noth...

Re: DeepSeek makes the V4 Pro price discount permanent

#516

Post by yogthos »

revolvingthrow wrote: post #288

Amusing that just when the big three AI providers from US raise prices significantly, even for the mini models, you’ve got a Chinese model slashing their already-cheap offer by 75%. Not to mention you can run this model on your own hardware, although admittedly even the flash stretches the meaning of local for individual people.

I imagine electricity costs being a third of what they are in the US in China has a lot to do with it.

Re: DeepSeek makes the V4 Pro price discount permanent

#517

Post by 1over137 »

rootusrootus wrote: post #478
dietr1ch wrote: post #442

Earlier quoted context omitted.

For most people Apple's main selling point is about showing off the cute devices and battery life, but that's not going to play a role when users are free to choose the tool that will call the models.

I’ve never seen anyone show off an iPhone. What a weird take.

Really? What country are you in?

Re: DeepSeek makes the V4 Pro price discount permanent

#518

Post by CuriouslyC »

roncesvalles wrote: post #455
DaSHacka wrote: post #454

Earlier quoted context omitted.

> attracting the best minds around the world to a free and stable society. China is comprised of ~91.5% ethnically Chinese citizens. [0] > Tump voters threw it all away because they couldn't stand non-whites coming to America and doing better than old stock Americans. The U.S. is more diverse than it's ever been [1], and under Trump we're still below the deportations of Obama's terms. Sounds like open-borders immigra…

You're conflating Mexican border hoppers with skilled immigrants. I'm absolutely opposed to illegal immigration and have a more extreme position on how to deal with it than most Americans. What I'm irked by are Trump's attacks on legal immigration and the general worsening of the environment. ICE's kidnappings, the 100k H-1B fee, and the recent Green Card thing have deeply eroded America's attractiveness to legal imm…

H1-Bs are the most egregious example, because they're 100% used as a way to undercut/replace American talent. The irony is that the typical border hopper is working jobs Americans don't want, for wages Americans wouldn't take, and they keep a low profile to avoid getting deported.

The desire to be influencers isn't as boneheaded as you think, in a future where AI is solving the hardest technical challenges, the ability to get attention and create community is the last frontier. Influencers and salesmen will be eating good when scientists and engineers are derelict.

Re: DeepSeek makes the V4 Pro price discount permanent

#519

Post by HDBaseT »

azinman2 wrote: post #513
HDBaseT wrote: post #505

Earlier quoted context omitted.

I'm not entirely sold on this idea, open source models aren't really hurting Deepseek or Qwens bottom line. 99.99% of people cannot run these models on their own hardware, they are forced to rent it from someone. That someone is almost always the big China players themselves anyways.

First, there’s manyyyy model inference providers out there world wide. Just look at open router. Second, it’s well known in SV that most startups are using Chinese models because they have access to the weights… and that makes it far cheaper. Why else is Qwen now having cloud-only models?

There is plenty of other inference providers, but tell me, who is the cheapest?

Model - Deepseek V4 Pro

CHEAPEST PROVIDER: Provider: Deepseek Input Price - $0.435/M tokens Output Price - $0.87/M tokens Cache Read - $0.003625/M tokens

SECOND CHEAPEST: Provider: deepinfra Input Price - $1.30/M tokens Output Price - $2.60/M tokens Cache Read - $0.10/M tokens

Deepinfra is almost 3x more expensive and they are using a fp4 model, with Max 16.4K output (vs 364K) and have significantly lower throughput!

Re: DeepSeek makes the V4 Pro price discount permanent

#520

Post by SmirkingRevenge »

DaSHacka wrote: post #454
roncesvalles wrote: post #402

Earlier quoted context omitted.

>Like there was something in the American DNA that was lacking in China and innovation would always need to happen here. There is (was): attracting the best minds around the world to a free and stable society. Trump voters threw it all away because they couldn't stand non-whites coming to America and doing better than old stock Americans.

> attracting the best minds around the world to a free and stable society. China is comprised of ~91.5% ethnically Chinese citizens. [0] > Tump voters threw it all away because they couldn't stand non-whites coming to America and doing better than old stock Americans. The U.S. is more diverse than it's ever been [1], and under Trump we're still below the deportations of Obama's terms. Sounds like open-borders immigra…

> The U.S. is more diverse than it's ever been [1], and under Trump we're still below the deportations of Obama's terms.

Ethnic diversity is neither really here nor there in terms of the measurable needs that immigration fulfills. Immigration keeps economic and population growth rates trending up. Having high skilled immigration to bolster science and research is nice, but it's still mainly about the growth.

Yea, Obama deported lots of people, but even then we still had net positive migration. Now under Trump, we have net negative migration for the first time in decades. The very public terror campaign waged by the Trump admin was in part to deter immigration in the first place.

> Sounds like open-borders immigration was never necessary in the first place, given that we're being beat by a country with a similar demographic skew that we had like 80 years ago.

1) Economic growth is possible with stagnating/declining population levels if you overcome those deficits with commensurate increases in productivity per capita. Otherwise, you're cooked.

2) The US is actually far more productive per capita than China - in fact, the US is one of the best in the world, as far as that goes.

With those points in mind, we can begin to see why China has an easier time growing economically with little immigration. The US has a much harder time doing the same. We need more population, since it's just harder to squeeze more productivity out of our already very productive workforce.

Once China achieves similar productivity levels, they will need to rely more on growing the population.

We were actually on track to catch up to China's population levels in a few of decades (thanks to immigration). So unless China successfully pivoted to mass immigration or expansionism, the US was likely to remain dominant - easily so - for the foreseeable future.

That's why the MAGA anti-immigration push is so tragically stupid and suicidal (if it persists). They're killing America's golden goose.

As an aside: I wish the "open borders" canard would die. We've never had open-borders immigration in recent history. Definitely not since 9/11. Not even under Biden. Border laws were enforced. Biden has the same apprehension rate at the border as both Trump and Obama.

Post reply on HN