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

Qwen 3.8 27B

huggingface.co

Post reply on HN
Print view View original

521–530 of 848 posts

Re: Qwen 3.8 27B

#521

Post by pbhjpbhj »

walrus01 wrote: post #510
simonw wrote: post #502

Earlier quoted context omitted.

It did a pretty outstanding job on that: https://tools.simonwillison.net/markdown-svg-renderer.html#u... This took 56 minutes! - but it didn't have exclusive use of LM Studio, I ran a few other test prompts while it was working. llm -m lmstudio/qwen/qwen3.8-27b 'Generate an SVG of a pelican equipment case'

That is pretty good! It's interesting it chose orange, because that's actually a somewhat rare color for pelican cases. I would say 95% of the ones going around the world right now are either black or FDE (flat dark earth) tan colored. There are some orange ones, for sure, if you google image search "orange pelican case", people use them for emergency/first aid supplies or electronics. But only the smallest ones are…

What are the chances that the colouring of actual pelicans is bleeding in to the prompt. Seems it could be that?

Re: Qwen 3.8 27B

#522

Post by walrus01 »

pbhjpbhj wrote: post #521
walrus01 wrote: post #510

Earlier quoted context omitted.

That is pretty good! It's interesting it chose orange, because that's actually a somewhat rare color for pelican cases. I would say 95% of the ones going around the world right now are either black or FDE (flat dark earth) tan colored. There are some orange ones, for sure, if you google image search "orange pelican case", people use them for emergency/first aid supplies or electronics. But only the smallest ones are…

What are the chances that the colouring of actual pelicans is bleeding in to the prompt. Seems it could be that?

Doubtful? Without knowing anything about how it was trained, if you google image search "pelican bird" they are predominantly white in color. Though the bill is certainly orange, different shades depending on camera and lighting conditions. The SVG attempt looks more like an attempt at the product photos if you search "pelican case orange".

Re: Qwen 3.8 27B

#523

Post by pbhjpbhj »

lacoolj wrote: post #516
ramon156 wrote: post #64

People will claim it's not comparable to Opus despite it beating the score. I'm not sure I disagree, but I'm also unsure whether I care. Most new models nowadays are "good enough". I cannot complain because I'd rather spend that time improving my prompts and docs. Opus might be a _slight bit better_ at picking up vague hints, but it's also extremely expensive, and I hit the 5 hour limit way too quick. I care a lot ab…

Grok 4.6 is a game-changer. I have yet to go back to other models after starting to use it. You just can't beat the price + output quality (even K3 is more expensive)

Supporting a far-right megalomaniac, whilst helping them to train their ML, and giving them all your data ... what could go wrong.

Re: Qwen 3.8 27B

#524

Post by InsideOutSanta »

satvikpendem wrote: post #485
gpm wrote: post #459

Earlier quoted context omitted.

I think it's very clear that someone who has checked out all the models but the one that called itself mechahitler and is explicitly being fine tuned to support far-right politics is making the choice for reasons other than performance and cost. It's not like all the other models even had plausible claims to those metrics.

As I said to a dead reply, for coding all of that is immaterial, as long as it codes well then that's all that matters to most people, except it seems those who have an idelogical issue in which case the other model companies also have issues.

I don't care how good Grok is, it will be a cold day in hell when I give a guy who goes to AfD rallies, instrumentally supported the current Ebola outbreak, and gives the Hitler salute on stage (don't try to tell me he did not, I saw the video) any money I'm not absolutely forced to.

And yes, OpenAI and Anthropic are highly questionable, too, and so I've ended my OpenAI subscription and I'm moving from Claude to other providers. But even people who do not do that are absolutely justified in drawing a line at supporting Musk, because there's a difference between mere normal shittiess and Musk.

Re: Qwen 3.8 27B

#525

Post by hannofcart »

Balinares wrote: post #248

I wonder if Anthropic and OpenAI possibly missed the window to go public. A 27B open-weight model trading blows with the SOTA from just half a year ago is not great news for trillion-dollar investments...

Another way of rephrasing that though is that the public (including large pension funds, retirees, etc.) just dodged a bullet and aren't left holding the bag.

Nevertheless, I doubt anyone's going to emerge unscathed when the valuations start falling though. Vast sections of the American stock market will be wiped out yet and with it the global economy.

Re: Qwen 3.8 27B

#526

Post by jdgoesmarching »

dannyw wrote: post #296
Balinares wrote: post #248

I wonder if Anthropic and OpenAI possibly missed the window to go public. A 27B open-weight model trading blows with the SOTA from just half a year ago is not great news for trillion-dollar investments...

Trading blows in some benchmarks is a bit exaggerated. If you try the model, `xhigh` is basically feels like the `max` mode (i.e. massive thinker and extremely presistent), and the amount of world knowledge and intent understanding is nowhere close to an Opus class model even from 6mo ago. It's still very useful, and it'll probably displace a good bit of API spend; but it's not really "trading blows with SOTA from ju…

Lacking world knowledge is fine for me, I rarely want to rely on the model’s training anyway when there are plenty of great search options to integrate with.

Re: Qwen 3.8 27B

#528

Post by distrill »

simonw wrote: post #419

Absolutely the best pelican I've seen from a model that runs on my laptop: https://tools.simonwillison.net/markdown-svg-renderer#url=ht... Bicycle is the right shape. Pelican beak is excellent. Nice background. Most importantly, the pelican has one leg on each side of the bicycle - that's very rare. (No chain on this bicycle though - in the reasoning trace it says "already chainstay... skip chain detail; maybe a smal…

this sort of test surely only works when it's a brand new concept/prompt. like the exact same models trained newer data will surely get better at putting this together over time

Re: Qwen 3.8 27B

#529

Post by notnullorvoid »

simonw wrote: post #419

Absolutely the best pelican I've seen from a model that runs on my laptop: https://tools.simonwillison.net/markdown-svg-renderer#url=ht... Bicycle is the right shape. Pelican beak is excellent. Nice background. Most importantly, the pelican has one leg on each side of the bicycle - that's very rare. (No chain on this bicycle though - in the reasoning trace it says "already chainstay... skip chain detail; maybe a smal…

I've done a few variations, I've been impressed with all of them. My favourite so far has been "Generate an SVG of a turtle flying a kite", result: https://imgur.com/a/bdKJPV4.

Some will say conflating flying and flying a kite is bad, but I like it.

Re: Qwen 3.8 27B

#530

Post by walrus01 »

Is it just me or is 3.6 27B Q8 K XL (Unsloth) holding up better in sustained token/s rate as the context fill increases over time? The token/s rate seems to be much higher for a time period deeper into context than previously seen.

At least as compared to 3.6 27B in the same quantization.

Post reply on HN