>>103996726
I'm gonna coom

>>103996726
Buy an ad.

>>103996726
>Deepseek-r1-Zero is the most uncensored model
LOL, no. It's even more pozzed then OpenAI.

>>103997088
This is because r1 knows basic human decency.

>>103996739
Can you make it give out the info you ask without formatting it like a gay reddit post?

>>103997165
Yes, that's just its internal monologue while it thinks. It will give a single "final answer" at the end when it's done.

>>103996739
problem is this is based on english data and the angloliberal consensus

>>103996739
>>103996726
Can it tell a man from a woman.

This is the real test of whether the model is grounded in reality or if its got built in "safety" protection for the castrated class in the west

>reddittards think tianamen square is test of censorship
Ask it about kikes

>>103996726
>>103996739
>>103997088
>words words words
>they made a leftoid into an ai personality
Lol this is so embarrassing for chinks after all the shilling

>>103996726
Did you know that the Tiananmen square massacre stared because african exchange students were raping chinese girls in college; and after weeks of complaining, the government didn't do anything so the chinese students had to take the problem in their own hands, and the communist goverment did everything thing to protect those niggers even putting tanks on the streets?

is deepseek the best coding AI now?

>>103996726
I'm too retarded to understand how to use hyperbolic to access r1-zero.
Please explain how to.

>>103999332
I tried it on my autohotkey scripts and it failed badly. Its unable to understand what the code does. My goto has been Google studio tbqh and thats been solid so far. O1 and also Grok does decent job of understanding it as well, but those are limited.

>>103999619
not zero

>>103999619
kek, literally everyone did this. fairs fair, ching chongs made a good AI and released it for free.
i think everyone including chinks knows what happened on 4th june 1989 by now. the funny thing is they're still trying this hard to keep it quiet after like 35 years and like 2 or 3 generations have grown up.

>>103999619
That's obviously some filtering step applied by the API server on the output generated by the model. You can download the model and use it yourself without any restrictions.

It's insane how unlike anything we had so far R1 is. It's actually a shock how well it can portray complex characters. Local seriously lost, for the time being. Can't go back to Llama after this. Guess I'll paypig until we get something comparable in the 70B range.

>>103999715
just use the distilled version locally

>>103999715
Also the full version is just 37B active parameters, you can run that locally

>>103999729
nowhere near as good barely better than the bases they're tuned on

I'd just be happy if those kike companies that invested trillions get their shit pushed in by a free chink LLM.
I don't like chinks, but I fucking despise jews. We should absolutely get behind this as a board.

>>103999755
assuming you also have 768gb of ram to store the inactive weights in

>>103999772
half that with quantization. it's doable and not even crazy expensive.

>>103999729
>>103999755
You are both retarded. The distills don't hold a candle to the real thing, they're more their parent models than R1, and I'm not about to keep swapping the model from SSD to get 0.0001 t/s when I can get near-instantaneous responses for cents.

>>103999803
You don't run it out of a ssd, you run it on the CPU. You could probably buy a rig to do this for less than $3k, but even if you go all in and build a current CPU and spend $10k on it, that's still very cheap for essentially o1 at home.

If it's open source why does it cost money to use?

Can I just not run it locally?

>>103999845
If you have the hardware sure.

>>103999845
compute and electricity costs

you can if you have more than 512gb of ram, the smaller versions aren't worth it

>>103999845
it costs money?

>>103999865
read the tweet

>>103999828
$10K will get me one hundred million tokens. How long do you think that will last me? Longer than the hardware would, I'd say.

>>103999940
i'm just using their free version idk

>>103999944
k own nothing and be happy then

>>103999944
The fact that you can get your own o1 running locally on your own hardware for $xx,xxx prices is kind of nuts tee bee aych

I don't think a lot of people have fully digested the implications of this.

>>104000262
Oh, it definitely is insane, the kinda thing I wouldn't have imagined until recently. Still, at that price point I'm actually better off going with the cheapest API that won't harvest too much personal data. The moment we get the same quality in a model one can reasonably run on a home PC, I'll be right back to local hosting though.

>>103996739
China should just come clean about Tiananmen Square. Xi Jinping should come up on the stage with a big picture of Tank Man and say "We squashed this nasty little malcontent because he deserved it. If you don't like it, flee to the West or we'll do the same to you", while a phonk remix of the National Anthem plays. Within a week, every nationalist in China will be spamming tank emojis all over the internet, domestic sales of tank merchandise will skyrocket, and China's enemies will lose one of their best pieces of ammunition.

>>104001175
I was just thinking about this! Every Deepseek thread always says something about
> but what about 1989 or Tiananmen Square
Might as well accept it. Maybe there is a translation that means
> You Can't Make an Omelette without breaking a few eggs

>>103996726
Will it generate smut?

>>103999845
It's hueg, so it's likely cheaper to let someone else host it and pay them per tokens instead. The cool thing is that you have the option to run it if you have the hardware, unlike models from OpenAI or Anthropic.

>>103999572
imagine admitting you're such a shitter that you use google studio for anything