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

Qwen 3.8

twitter.com

Post reply on HN
Print view View original

601–610 of 793 posts

Re: Qwen 3.8

#601

Post by jcelerier »

akoboldfrying wrote: post #498
windsignaling wrote: post #487

Earlier quoted context omitted.

There is clearly an anti-China bias here. Show me comments demonstrating the same level of distrust against Google for open-sourcing projects like Tensorflow, Kubernetes, Flutter, Chromium, etc.

I'll supply such a comment: Any software open sourced by any for-profit company, including Google, is a calculated move ultimately intended to increase their bottom line, and it's naive to think otherwise.

Isn't it borderline illegal for publicly traded companies to not try to maximize profit? "interest of the company" in theory but in practice

Re: Qwen 3.8

#602

Post by dackdel »

gardnr wrote: post #6
adrian_b wrote: post #4

I assume that this announcement has been prompted by that of Moonshot AI, which has just announced a 2.8T parameter open-weights LLM, Kimi K3, to be published on Huggingface by 27 July. Now the response of Alibaba is that they will also publish soon a big open weights LLM, the 2.4T parameter Qwen 3.8. I wonder if Alibaba has always planned to make this big LLM open weights, or they have chosen to do this now, to bett…

It's hard to say what their motivation is. The Chinese firms seem to be working hard to commoditize intelligence which may be the most effective way to debase American frontier labs. And yeah: it also happens to be really good for humanity.

its hard to say what their motivation is, if you live under a rock and have severe brain damage.

Re: Qwen 3.8

#604

Post by MintPaw »

tymscar wrote: post #565
benjaminoakes wrote: post #564

Earlier quoted context omitted.

In my opinion, simonw shouldn't have to play those games

I find it so strange when someone comments something like this. Nobody thinks this is okay or that Simon should have to play those games. The point of the reply was that while this is stupid, the solution is trivial. I’m genuinely very curious what the point of comments like this is. I am not joking, I want to understand. I see examples like it 50 times a week in random places, and nobody walks me through their menta…

I'm not sure what the situation is here, but on many sites making alt account to avoid bans can result in a lifetime ban. A high profile figure who's been banned from a service advertising that they still have access to it might be a risk.

Also, these companies (or data brokers) likely share ban and alt account lists, ban evading could hurt the reputation of all your accounts.

These are some pragmatic reasons not to evade bans.

Re: Qwen 3.8

#605

Post by adverbly »

cnhwl wrote: post #592
adithyassekhar wrote: post #587

Earlier quoted context omitted.

Unfortunately this is a US centric site, for a US based investor, who themselves and probably most commenters as well who are highly paid silicon valley people who have stakes in ai companies on their side of the pond. Not to mention the political unrest and fear of losing the technological superiority which they once earned but now is trying so hard to hold on to through legally grey monopolistic practices. You dare…

Rankings have never been something I’ve cared about. Just as with the smear campaigns against China, there is far too much noise in the world. We simply hope that the geeks who are genuinely committed to making the world a better place can unite and focus on getting things done, rather than being brainwashed by ideology and social media.

> committed to making the world a better place can unite and focus on getting things done

The problem is that nobody can agree on what "better" is, so it is difficult to unite.

And in the meantime, people who don't care about "uniting" or "better" just carry on with whatever their goal is(usually just to make as much money as possible), and that sometimes ends up on net making the world a worse place.

This is a legitimate systematic problem that many of us are concerned about solving. It could easily be more important than the technical problems that geeks such as myself would love to be able to focus on.

Re: Qwen 3.8

#606

Post by bigyabai »

villish wrote: post #599
adithyassekhar wrote: post #587

Earlier quoted context omitted.

Unfortunately this is a US centric site, for a US based investor, who themselves and probably most commenters as well who are highly paid silicon valley people who have stakes in ai companies on their side of the pond. Not to mention the political unrest and fear of losing the technological superiority which they once earned but now is trying so hard to hold on to through legally grey monopolistic practices. You dare…

Legally gray practices? Such as what? Old politicians who barely understand how to use e-mail had a panic attack due to Anthropics marketing about Mythos and being told that Chinese models were "attacking" (distillation) US frontier labs to train theirs. They pull the only lever they could in an attempt to stop it, but again, they don't understand how any of it works. The export ban lasted a few weeks yet everyone ke…

> I personally think it is inevitable that China and the US stops producing frontier level open models when they become too valuable.

This reads like a paradox, though. If the value of a FOSS model comes from pushing the local frontier, then FOSS releases will be even more attractive in a world with highly capable AI services. The underdog researchers will be motivated to train smaller LLMs that close the gap with the most newest architectures and improvements, democratizing those same benefits to local models. There's not really anything that stops them, besides finding an investor.

The US already has top secret AI technology like the NRO's Sentient. The danger is less the architecture, and more the data that it has access to and the technology that it interfaces with. Most people aren't entirely convinced that AI will ever be "too valuable" in the first place.

Re: Qwen 3.8

#607

Post by neya »

gardnr wrote: post #6
adrian_b wrote: post #4

I assume that this announcement has been prompted by that of Moonshot AI, which has just announced a 2.8T parameter open-weights LLM, Kimi K3, to be published on Huggingface by 27 July. Now the response of Alibaba is that they will also publish soon a big open weights LLM, the 2.4T parameter Qwen 3.8. I wonder if Alibaba has always planned to make this big LLM open weights, or they have chosen to do this now, to bett…

It's hard to say what their motivation is. The Chinese firms seem to be working hard to commoditize intelligence which may be the most effective way to debase American frontier labs. And yeah: it also happens to be really good for humanity.

In case of American labs, always follow the money. In case of Chinese labs, always follow the IP.

Re: Qwen 3.8

#608

Post by nicman23 »

kelnos wrote: post #324
gardnr wrote: post #6

Earlier quoted context omitted.

It's hard to say what their motivation is. The Chinese firms seem to be working hard to commoditize intelligence which may be the most effective way to debase American frontier labs. And yeah: it also happens to be really good for humanity.

> It's hard to say what their motivation is. Feels pretty easy to me. They want to turn LLMs into a commodity, and watch the US AI labs crash and burn. There will still be plenty of customers who will pay them to host the models and run inference, even if the weights are open and others can offer competing products. (If necessary, the Chinese government can ban use of foreign inference services by Chinese citizens an…

to be honest i think this is a open source vs close source software.

some companies chose to have the benefits of one or the other

Re: Qwen 3.8

#609

Post by villish »

bigyabai wrote: post #606
villish wrote: post #599

Earlier quoted context omitted.

Legally gray practices? Such as what? Old politicians who barely understand how to use e-mail had a panic attack due to Anthropics marketing about Mythos and being told that Chinese models were "attacking" (distillation) US frontier labs to train theirs. They pull the only lever they could in an attempt to stop it, but again, they don't understand how any of it works. The export ban lasted a few weeks yet everyone ke…

> I personally think it is inevitable that China and the US stops producing frontier level open models when they become too valuable. This reads like a paradox, though. If the value of a FOSS model comes from pushing the local frontier, then FOSS releases will be even more attractive in a world with highly capable AI services. The underdog researchers will be motivated to train smaller LLMs that close the gap with th…

> Most people aren't entirely convinced that AI will ever be "too valuable" in the first place.

There will be a generational leap that makes current LLMs obsolete in every way. Let's say China gets there first, why would the CCP let it be released publicly? That is an insanely valuable advantage in everything from war fighting to economics and more.

Humans for millennia have used technological advances to make better weapons. AI will be no different, unfortunately. Wars will be fought with drones and artificial intelligence from now on.

Re: Qwen 3.8

#610

Post by nicman23 »

embedding-shape wrote: post #224
androiddrew wrote: post #172

Earlier quoted context omitted.

Curious, do you find the 3.5 120B sized MoE works better than the dense 3.6 27B?

Yeah, Qwen3.5-122B-A10B-NVFP4 produces better responses than Qwen3.6-27B-NVFP4 (both from unsloth), but I'm mostly using them for programming in various ways, mostly Rust, Clojure, Python and JavaScript, and some translations tasks, but not much more than that, so YMMV. Edit: as a concrete example, I'm working on a "optimization framework via agent harness" right now, Qwen3.6-27B-NVFP4 is often unable to actually com…

dense small models do not like quantization. i find 27b fp8 to be smarter albeit less knowledgable versus the 122B
Post reply on HN