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

Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

qwen.ai

Post reply on HN
Print view View original

341–350 of 400 posts

Re: Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

#341

Post by 1dom »

robot_jesus wrote: post #109
culi wrote: post #99

Earlier quoted context omitted.

I think that's an overgeneralization. We've seen all the American models be closed and proprietary from the start. Meanwhile the non-American (especially the Chinese ones) have been open since the start. In fact they often go the opposite direction. Many Chinese models started off proprietary and then were later opened up (like many of the larger Qwen models)

> We've seen all the American models be closed and proprietary from the start What about Gemma and Llama and gpt-oss, not to mention lots of smaller/specialized models from Nvidia and others? I would never argue that China isn't ahead in the open weights game, of course, but it's not like it's "all" American models by any stretch.

It's all openwashing, all of the ones you listed at somepoint have expressed how important and valuable open weights and locally usable models are. Every single one of them has then increasingly focused and pushed closed, proprietary or cloud usable only options since saying/doing that.

I'm annoyed at myself, because I thought/hoped/praised chinese AI when they were opening up as Llama was closing, but Qwen looks to be doing the same playbook here as Llama/Meta, Gemma/Google and OpenAI/gpt-oss.

Re: Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

#342

Post by mettamage »

mikenew wrote: post #261
alex7o wrote: post #71

Ok I find it funny that people compare models and are like, opus 4.7 is SOTA and is much better etc, but I have used glm 5.1 (I assume this comes form them training on both opus and codex) for things opus couldn't do and have seen it make better code, haven't tried the qwen max series but I have seen the local 122b model do smarter more correct things based on docs than opus so yes benchmarks are one thing but realit…

GLM 5.1 was the model that made me feel like the Chinese models had truly caught up. I cancelled my Claude Max subscription and genuinely have not missed it at all. Some people seem to agree and some don't, but I think that indicates we're just down to your specific domain and usage patterns rather than the SOTA models being objectively better like they clearly used to be.

Hmm

Will try it out. Thanks for sharing!

Re: Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

#343

Post by jpfromlondon »

paprikanotfound wrote: post #143
jpfromlondon wrote: post #75

Earlier quoted context omitted.

anecdotally the quality of output isn't significantly different, the speed seems to be what you're really paying for, and since the alternative is free I'll stick to local.

What are the best models to run locally?

right now Gemma 4 and Qwen 3.6, I've found the latter to have the slight edge but your results may vary.

Re: Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

#344

Post by Mashimo »

Mashimo wrote: post #169
Moosdijk wrote: post #164

Earlier quoted context omitted.

I wonder why glm is viewed so positively. Every time I try to build something with it, the output is worse than other models I use (Gemini, Claude), it takes longer to reach an answer and plenty of times it gets stuck in a loop.

I have used GLM 4.7, 5 and 5.1 now for about 3 month via OpenCode harness and I don't remember it every being stuck in a loop. You have to keep it below ~100 000 token, else it gets funny in the head. I only use it for hobby projects though. Paid 3 EUR per month, that is not longer available though :( Not sure what I will choose end of month. Maybe OpenCode Go.

EDIT: Ok, now I tried GLM for the first time in the morning CET, and it was .. bad. The reasoning took 5 mintues for a very very small .html file going around in circles.

Evening CET experience for me is super smooth.

Re: Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

#345

Post by dTal »

thesz wrote: post #224
dTal wrote: post #137

Earlier quoted context omitted.

None of that is true, at least in theory. You can trivially change layer size simply by adding extra columns initialized as 0, effectively embedding your smaller network in a larger network. You can add layers in a similar way, and in fact LLMs are surprisingly robust to having layers added and removed - you can sometimes actually improve performance simply by duplicating some middle layers[0]. Tokenization is probab…

You took a simple path, embedding smaller into larger. What if you need to reduce number of layers and/or width of hidden layers? How will you embed larger into smaller? As for the "addition of same layers" - would the process of "layers to add" selection be considered training? What if you still have to obtain the best result possible for given coefficient/tokenization budget? I think that my comment express general…

The general case is that our own current relative ignorance on the best way to use and adapt pretrained weights is a short-lived anomaly caused by an abundance of funding to train models from scratch, a rapid evolution of training strategies and architectures, and a mad rush to ship hot new LLMs as fast as possible. But even as it is, the things you mentioned are not impossible, they are easy, and we are only going to get better at them.

>What if you need to reduce number of layers

Delete some.

> and/or width of hidden layers?

Randomly drop x% of parameters. No doubt there are better methods that entail distillation but this works.

> would the process of "layers to add" selection be considered training?

Er, no?

> What if you still have to obtain the best result possible for given coefficient/tokenization budget?

We don't know how to get "the best result possible", or even how to define such a thing. We only know how to throw compute at an existing network to get a "better" network, with diminishing returns. Re-using existing weights lowers the amount of compute you need to get to level X.

Re: Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

#346

Post by chillfox »

blurbleblurble wrote: post #321
alex7o wrote: post #71

Ok I find it funny that people compare models and are like, opus 4.7 is SOTA and is much better etc, but I have used glm 5.1 (I assume this comes form them training on both opus and codex) for things opus couldn't do and have seen it make better code, haven't tried the qwen max series but I have seen the local 122b model do smarter more correct things based on docs than opus so yes benchmarks are one thing but realit…

Opus 4.6 was incredible but Opus 4.7 is genuinely frustrating to me so far. It's really sharp but can be so lazy. It's constantly telling me that we should save this for tomorrow, that it's time for bed (in the middle of the day), and very often quite sloppy and bold in its action. These adjustments are getting old. The next crop of open models seems ready to practically replace the big ones as sharp orchestrator age…

I have never seen a model be “lazy” before (I have seen them go for minimal change). I have been using the models through the api with various agents and no custom system prompt.

So I am curious, how do people get these lazy outputs?

Is it by having one of those custom system prompts that basically tells the model to be disrespectful?

Or is it free tier?

Cheap plans?

Re: Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

#347

Post by chillfox »

Moosdijk wrote: post #164
alex7o wrote: post #71

Ok I find it funny that people compare models and are like, opus 4.7 is SOTA and is much better etc, but I have used glm 5.1 (I assume this comes form them training on both opus and codex) for things opus couldn't do and have seen it make better code, haven't tried the qwen max series but I have seen the local 122b model do smarter more correct things based on docs than opus so yes benchmarks are one thing but realit…

I wonder why glm is viewed so positively. Every time I try to build something with it, the output is worse than other models I use (Gemini, Claude), it takes longer to reach an answer and plenty of times it gets stuck in a loop.

GLM is the first open source model that actually worked for me, where I found the output ok.

And yes, sonnet/opus is better and what I use daily. But I wouldn’t be that upset if I had to drop down to GLM.

Re: Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

#348

Post by szundi »

blurbleblurble wrote: post #321
alex7o wrote: post #71

Ok I find it funny that people compare models and are like, opus 4.7 is SOTA and is much better etc, but I have used glm 5.1 (I assume this comes form them training on both opus and codex) for things opus couldn't do and have seen it make better code, haven't tried the qwen max series but I have seen the local 122b model do smarter more correct things based on docs than opus so yes benchmarks are one thing but realit…

Opus 4.6 was incredible but Opus 4.7 is genuinely frustrating to me so far. It's really sharp but can be so lazy. It's constantly telling me that we should save this for tomorrow, that it's time for bed (in the middle of the day), and very often quite sloppy and bold in its action. These adjustments are getting old. The next crop of open models seems ready to practically replace the big ones as sharp orchestrator age…

[dead]

Re: Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

#349

Post by enraged_camel »

chillfox wrote: post #346
blurbleblurble wrote: post #321

Earlier quoted context omitted.

Opus 4.6 was incredible but Opus 4.7 is genuinely frustrating to me so far. It's really sharp but can be so lazy. It's constantly telling me that we should save this for tomorrow, that it's time for bed (in the middle of the day), and very often quite sloppy and bold in its action. These adjustments are getting old. The next crop of open models seems ready to practically replace the big ones as sharp orchestrator age…

I have never seen a model be “lazy” before (I have seen them go for minimal change). I have been using the models through the api with various agents and no custom system prompt. So I am curious, how do people get these lazy outputs? Is it by having one of those custom system prompts that basically tells the model to be disrespectful? Or is it free tier? Cheap plans?

I have seen some people complain about a new tendency where it can suggest wrapping up the current task even though it isn't done yet. I haven't seen it myself though.

Re: Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

#350

Post by elorant »

Barrin92 wrote: post #303
elorant wrote: post #80

Earlier quoted context omitted.

This is obviously a strategic move at a national level. Keep publishing competing free models to erode the moat western companies could have with their proprietary models. As long as the narrative serves China there will be no turn to proprietary models.

>This is obviously a strategic move at a national level. no it isn't. That's the kind of thing people say who've never worked in the Chinese software ecosystem. It's how the Chinese internet has worked for 20+ years. The Chinese market is so large and competition is so rabid that every company basically throws as much free stuff at consumers as they can to gain users. Entrepreneurs don't think about "grand strategic…

If this was true then they’d build services around those models and provide those for free or vastly cheaper than western competition. But that’s not what they’re doing. Instead they’re giving away the entire model for free. And by the way, Qwen isn’t build from some random entrepreneur who’s trying to solve the cold start problem, but from Alibaba which is a fucking behemoth. And surprisingly of course none of these models answer uncomfortable questions about China’s past. Because sure enough, the first thing any entrepreneur would think is to protect their government and their history. Sure, happens all the time, no state interference here, move on.
Post reply on HN