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

421–430 of 848 posts

Re: Qwen 3.8 27B

#421

Post by simonw »

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…

For anyone who followed yesterday's Gemini 3.7 Flash pelican which rendered in Safari but not in Firefox or Chrome... https://news.ycombinator.com/item?id=49289112#49290012. - that turned out to be my fault, not the model.

My SVG rendering software was stripping some attributes. Here's the Gemini 3.7 Flash pelicans in the fixed renderer: https://tools.simonwillison.net/markdown-svg-renderer.html#u...

Re: Qwen 3.8 27B

#422

Post by sheepscreek »

anana_ wrote: post #17

Monstrous benchmarks! Hoping it is not benchmaxxed.

I thought the same. But why claim something so shocking when it can easily be discredited and puts your reputation at risk? If they’re claiming Opus 4.6 level, I expect it to at least match Sonnet 4.6.

Re: Qwen 3.8 27B

#423

Post by pulse7 »

qiine wrote: post #396
hypfer wrote: post #109

Since it might be helpful to some, here's my current commandline for llama.cpp running on an RTX 4090 with my monitor moved to the iGPU to free up all of its VRAM. llama-server -m Qwen3.8-27B-IQ4_NL.gguf --mmproj mmproj-BF16.gguf -c 170000 --parallel 1 -ngl -1 --cache-type-k q8_0 --cache-type-v q8_0 -b 1024 -ub 512 --flash-attn on --no-context-shift --no-mmproj-offload --spec-type draft-mtp --spec-draft-n-max 5 --spe…

what --no-context-shift is for ?

"--context-shift, --no-context-shift ... whether to use context shift on infinite text generation (default: disabled)"

From: https://github.com/ggml-org/llama.cpp/blob/master/tools/serv...

Re: Qwen 3.8 27B

#424

Post by fragmede »

beembeem wrote: post #374
jacquesm wrote: post #340

Earlier quoted context omitted.

I absolutely love this comment. I wished there was a website where people would post their working command lines as well as what hardware they are using to run that stuff on + tokens / sec prefill + gen.

What stackoverflow should have become.

still could! there's the network site https://ai.stackexchange.com/ just need people to ask questions and to downrank accepted answers older than a year (that might not be possible though).

Re: Qwen 3.8 27B

#425

Post by cloudengineer94 »

Been trying out Qwen3.8-27B-Q5_K_S_20GB and it's quite interesting it's behaving very well.

Going to give it some coding tasks and see how it goes.

We been eating good at LocalLlama this week.

Re: Qwen 3.8 27B

#426

Post by fragmede »

bilekas wrote: post #380
formerly_proven wrote: post #260

Earlier quoted context omitted.

> I sure hope my boss doesn't think he built my work! Most managers do though?

I've worked with a good few types of managers of the years, none would take ownership of my work. The end result? Sure, for shareholders/this managers that is fine. But that's not the same as me being needed to be on a call with integration teams and having to run the call. It's just a layer of abstraction for management. I own the creation side of the product. Who created it is still me.

Did Steve Jobs create the iPhone?

Re: Qwen 3.8 27B

#427

Post by throw310822 »

kristjansson wrote: post #204
ramon156 wrote: post #69

Earlier quoted context omitted.

hence the "former deepseek v4 pro". I tried it out this morning and have had no complaints. I already liked glm 5.2

> Deepseek v4 pro 0813 Which itself released yesterday? You're writing, reading, and evaluating enough software in a ~36 hour period to form, reject, and form another opinion about which model makes better _architectural_ choices?

It's the singularity, baby.

Re: Qwen 3.8 27B

#428

Post by acka »

CamperBob2 wrote: post #146
bmitc wrote: post #138

Earlier quoted context omitted.

Lol at that command. Why is this stuff so hard to run locally? I've spent a few days trying to figure it all out and haven't been able to. LM Studio doesn't work behind proxies. Ollama is confusing and doesn't seem to support Qwen3? And Llama.cpp is your command. I just want to run ` ` with some default parameters set and for it to run locally.

I never install this stuff manually anymore. Just tell your LLM of choice to download model X from URL Y, build the latest inference engine of choice E, and then create batch files or shell scripts to run instruct and/or reasoning models in accordance with instructions at URL Z.

While you probably have the audacity to check the commands and code that your favorite LLM generates to complete this task, i'm pretty much certain that at some point in time, due to this pervasive mindset of "just let an AI agent do it," we'll have a front page news story of someone who had their family photos on their NAS uploaded to some public space, their security cameras connected to a "live porn for lolz" webring, and perhaps even their crypto wallets and bank accounts pilfered, all because their favorite LLM got exploited by some prompt injection attack while trying to be helpful doing system administration tasks.

I remember there was a short story in BYTE Magazine about a similar kind of scenario way back when, I think at least 30 years ago, long before LLMs and AI agents became a reality.

Re: Qwen 3.8 27B

#429

Post by Casteil »

kzrdude wrote: post #398
isomorphic wrote: post #367

Earlier quoted context omitted.

"inclined to overthinking," holy cats you're not kidding! On a Mac mini M4 Pro 64GB I prompted it with "svg owl" and it thought for 17m12s, outputting 36.3KiB of thinking chatter. It did end up producing a 20.2KiB HTML+JS+SVG file with a very nice owl, including cursor-tracking animation, but it ran for more than a half hour! The MoE models are stupendously faster.

I think that's called test-time scaling i.e using more tokens at infer time to squeeze out higher model performance. That's must be part of the explanation for good benchmark results.

Yep.. it's pretty obnoxious for real-world use with the default 'xhigh' thinking. Ridiculous amount of "Wait, actually.." which might help for complex coding tasks but makes it unbearable for general purpose use.

Re: Qwen 3.8 27B

#430

Post by mrshu »

Casteil wrote: post #82

One thing a lot of people don't seem to factor when hyping Qwen is how much models like this tend to 'overthink' with seemingly endless 'second guessing'. 3.8 seems no different from what I've tried thus far. As capable as it is, it's hard to justify using it when a competing model (e.g. Gemma4:26b-a3b) can consistently achieve the same or similar response with only 1/10th as many 'thinking' tokens, achieve much high…

This is why the "reduce thinking" approach of BottleCapAI is both quite interesting and very useful in practice:

https://huggingface.co/bottlecapai/ThinkingCap-Qwen3.6-27B

Post reply on HN