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

DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

artificialanalysis.ai

Post reply on HN
Print view View original

271–280 of 343 posts

Re: DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

#271

Post by bethekind »

cmrdporcupine wrote: post #213
ycui7 wrote: post #207

Earlier quoted context omitted.

the rational in one’s mind is similar to buying expensive supercar but no driving it daily. owning a few GPUs is a lot cheaper than supercars.

I dunno. I bought the Spark in January and it has led indirectly to paid work. I don't use it for local inference so much. I use it to learn. I also use it as my daily driving Aarch64 development system. Aside it's also very cool what else can be done with unified GPU memory, once you realize you have it...

Learn model deployment, batching, all the AI inference related things?

Re: DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

#272

Post by johnny_rico »

xXSLAYERXx wrote: post #258
dpacmittal wrote: post #91

[flagged]

You're clearly not an American and they will never catch up on hardware. I'm mostly curious why you are so excited for China to "crush" American companies?

Incursions into other nations "just because"? Economic wars, and support for foreign governments or armed groups that caused harm or instability to the entire planet? You know? Planet Earth?

Sanctions and trade policies that hurt ordinary people? Cultural dominance through Hollywood (Holy Wood, again with the sick ruling elite and their witchcraft crap), social media, brands, the poisoning of food and water on global scale?

Etc? Etc?

Most people don't hate "Americans" or their companies, they are just against your ruling elite. Live one day outside the united states and understand how the planet works for the rest of mankind and you might start to grasp on reality.

And yeah, China will catch up with hardware pretty damn soon. Jesus, you guys bought your radios from Japan not 50 years ago. Sorry, can't dominate the world's economy forever. Never happened in history.

No matter how much black magic, dark rituals, sick crap and sacrifices people on the top throws at the wall.

We're humans, after all.

So yeah, it is not against America itself or American companies, but against the ruling class. So naturally , people get excited about China entering into the match!

3...2...1... Fight!

Re: DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

#273

Post by dgellow »

UltraSane wrote: post #223
hgoel wrote: post #215

Earlier quoted context omitted.

By ignoring the 1st amendment? The US has a tendency of ignoring its constitution whenever it's convenient.

Don't be glib, the 1st amendment is taken VERY seriously.

Im not sure if you’re joking or not.

- the US strictly regulate cryptography https://en.wikipedia.org/wiki/Export_of_cryptography_from_th...

- some prime numbers are considered illegal https://en.wikipedia.org/wiki/Illegal_number#Illegal_primes

I don’t think banning open weight is that far fetch compared to those

Re: DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

#274

Post by shimman »

Saline9515 wrote: post #137
dpacmittal wrote: post #91

[flagged]

Good luck when the last non-Chinese frontier labs will have closed and the CCP will ask to stop sharing models open source.

As opposed to closed source American models where the federal government threatens your livelihood unless you build data centers to extend pax silica?

Re: DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

#275

Post by cherryteastain »

baq wrote: post #265

Performance per dollar is in the ‘too good to be true’ territory, what’s the deal here?

This is China's way of applying its modus operandi of '80% the quality at 20% the price' to LLMs. Aim is dumping the tokens on the world market like they did with other industries like shipbuilding, steel etc.

Re: DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

#277

Post by sourcecodeplz »

scosman wrote: post #85

So GLM 5.2/Gemini 3.6 level intelligence for $0.28/m output. And their updated Pro model coming soon.... Plus a size you can genuinely run at home: Unsloth lossless Q8 at 162GB.

i am scared for the PRO model

maybe it is Fable level

Re: DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

#278

Post by krull10 »

adrian_b wrote: post #72
vehemenz wrote: post #35

Earlier quoted context omitted.

This is a straightforward false equivalency. “Western” models do not censor in the same way, nor for the same reasons, that the Chinese models do. “Just as much” is not remotely plausible, yet it’s doing all the heavy lifting.

The Anthropic and OpenAI models are much more censored and in ways that directly prevent them to be useful, e.g. by refusing to reply to elementary questions of biology and chemistry. Any normal user is much more likely to ask questions to which the Anthropic and OpenAI models do not answer, than to ask questions about the modern Chinese history, to which a Chinese LLM will not answer.

I actually cancelled my max plan for Claude today due to this silliness. I can’t get Fable to answer a single question, even pure mathematical questions that have no bio/chem references. It was a totally worthless addition to my account…

Re: DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

#279

Post by VulgarExigency »

dominotw wrote: post #263
VulgarExigency wrote: post #162

Earlier quoted context omitted.

And what is their nefarious plan after we get "addicted to cheap shit"?

same plan as always. stop from producing native capablity and then use the dependency to your advantage like becoming un attackable. cheap shit is more powerful that a nuclear shield.

But why would "we" want to attack them anyway?

Re: DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

#280

Post by felixgallo »

stavros wrote: post #249
felixgallo wrote: post #77

Earlier quoted context omitted.

what plan are your 'fellow software engineers' using? I have a hard time even using up the Fable part of my allowance in a week of coding.

I'm on the Premium business plan and I can easily use up my entire week's allocation in two days of coding. I only use Claude for planning, too, the rest is done by Deepseek and GPT. Claude is very spendy.

Claude is about the least spendy thing a software engineer can use. If you can't afford $200 a month, you're probably not working professionally (which is ok, no value judgement, but then being on the premium business plan is a little strange).
Post reply on HN