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

191–200 of 848 posts

Re: Qwen 3.8 27B

#191

Post by mechagodzilla »

D4Ha wrote: post #133
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…

Do you find it useful or worthwhile to split a large LLM across two GPUs on a desktop? If you've tried it, what worked well and what didn't? I'm especially interested in mismatched VRAM setups, e.g. a 16 GB GPU + a 24 GB GPU. How much overhead did you see from inter-GPU transfers, and did the extra usable VRAM outweigh the performance hit?

Yes. I can split a model like this across 3 GPUs (a 1080 with 8GB and two Titan Vs with 12GB), and it's much faster than running it on 36 CPU cores. As long as it fits in aggregate VRAM, it seems very advantageous to do so.

Re: Qwen 3.8 27B

#192

Post by cyanydeez »

peri-cl wrote: post #27
KronisLV wrote: post #10

I hope really badly that we'll get a new 35B A3B or similar MoE model! I also miss the Qwen 3 Coder Next, which was 80B A3B, there are quite a few use cases where a non-dense model Also alternate link for viewing the images without signing in: https://xcancel.com/Alibaba_Qwen/status/2088280182356611304

Same here! Qwen3.6-35B-A3B is the only local model I've found that runs reasonably on my iGPU. Looks like me and and my noisily-wheezing laptop will be sitting out this upgrade.

yeah, that's the A3B part; going up to A5B would probably also feel comfortable.

on the 395+ AI MAX w/128GB, the A10B qwen 3.5 can do a lot of long running work if you don't need to baby sit it. deer-flow works well like that.

Re: Qwen 3.8 27B

#193

Post by danielhanchen »

satvikpendem wrote: post #172
nblgbg wrote: post #14

Earlier quoted context omitted.

Is there any advantage to using the model from Unsloth compared with https://huggingface.co/Qwen/Qwen3.8-27B-FP8 ?

Unsloth usually also fixes the models when they bork something, which always happens. For Gemma for example the tool calling wasn't working for the longest time.

That wasn't our problem right? Gemma officially updated tool calling which we adopted

Re: Qwen 3.8 27B

#194

Post by segmondy »

Almondsetat wrote: post #132

The $1500 Intel B70 with 32GB of VRAM can run this model at max context with good performance, btw. If you don't want to drop $5-10k for running DeepSeek this is your best budget option for local refactor/small scale dev help

You don't need $10k to run DeepSeek, I run it on a $1000 system.

Re: Qwen 3.8 27B

#195

Post by nullc »

mips_avatar wrote: post #186
D4Ha wrote: post #133

Earlier quoted context omitted.

Do you find it useful or worthwhile to split a large LLM across two GPUs on a desktop? If you've tried it, what worked well and what didn't? I'm especially interested in mismatched VRAM setups, e.g. a 16 GB GPU + a 24 GB GPU. How much overhead did you see from inter-GPU transfers, and did the extra usable VRAM outweigh the performance hit?

Depends on your pcie connection. If they're both x16 then it's pretty low overhead, x8 is ok, but x4 is too slow. Also it's a bit tricky getting an optimal setups with mismatched vram, I think you could probably still make use of the full vram if you're clever but it's trickier.

for layer parallelism (e.g. to get more vram) the bandwidth between layers is essentially nothing (like 16kb per token I think), so I don't think x4 would even be a problem!

Re: Qwen 3.8 27B

#196

Post by tinfoilhatter »

nubg wrote: post #165
hypfer wrote: post #124

Earlier quoted context omitted.

According to this shitty vibecoded thing "I" built https://hypfer.github.io/will-it-fit-llama-cpp/ (and I guess according to math too), FP16 K/V would give me something like 90k context at the same model quant, which doesn't really fit my usage. But maybe someone else has experience to share there

just to clarify. yes YOU built it. just because you used some tool doesn't mean the idea, prompting, reprompting, babysitting was not your creative input and effort. put differently, if you put a random person infront of whatever model you used (say, a 50yo receptionist at a pharmacy in india), they would not have been able to create that, because they would have lacked the motivation, idea, background knowledge, tas…

So if I hire an artist and am a motivated individual, have an idea for a painting, have background knowledge about paintings and have taste in paintings and can offer a critique of the painting as the artist paints it, then somehow I created the painting?

Absurd logic. The AI built the website.

Re: Qwen 3.8 27B

#197

Post by ericd »

arjie wrote: post #135
ericd wrote: post #105

Earlier quoted context omitted.

Was recently thinking about doing something similar, do you basically just have the qwens describe what they see for the flashes? Was considering adding a LoRa/vision head to Flash, but seems like it could take a while to get it right. If DSv4 Flash was multimodal, I’d probably be done model shopping for a while

Same, with a multimodal DSv4 Flash I would just stop paying attention to things. Very smart, and at 260 tok/s it's too fast to care about anything else. If you ever graft something like that I would love to hear about it. Yes, I have a very dumb flow. The harness has a describe_image tool that takes an image and a prompt and so DSv4 Flash uses it to get an idea of what it's looking at.

Yeah, I might just replicate what you're doing. Main issue right now is just finding spare vram to actually run another model in parallel... And yeah, if I train up a vision adapter somehow, I'll try to put it up/post about it, seems like we're getting the killer apps for local LLMs right now, where it's just feasible enough if you're enthusiastic enough to be a bit economically irrational, and just useful enough to sort of rationalize.

Re: Qwen 3.8 27B

#198

Post by danielhanchen »

zargon wrote: post #114
scrlk wrote: post #74

Earlier quoted context omitted.

Unsloth publishes KL divergence numbers which measures how much the quantised probability distribution changes vs unquantised: https://unsloth.ai/docs/models/qwen3.8#quantization-analysis It's a bit bare at the moment, I assume they are going to add further detail later (eg comparison to other quants), similar to their other releases.

KL divergence is nothing close to a replacement for benchmarks. As flawed as benchmarks are, KL divergence is a barely useful signal. The fact that Unsloth only just started publishing KL divergences shows how unserious the quantization space is.

Actually we do publish non KLD benchmarks - top-1% is better - for NVFP4 for eg we did MMLU Pro, GPQA, AIME 2025: https://unsloth.ai/docs/models/qwen3.6#nvfp4-benchmarks

Sometimes they're just slow and expensive, so we we KLD as a proxy measure and it's very high correlation (95%+)

Re: Qwen 3.8 27B

#199

Post by ycui7 »

redox99 wrote: post #143
NitpickLawyer wrote: post #34

Earlier quoted context omitted.

> Beats Opus 4.7 Max I'm a huge open model fan, and have used them since forever, even have daily drivers for on-prem dev, but no. They do not beat opus on real-world usage. Qwen models are impressively good for what they are, are "good enough" for plenty tasks, can be ran locally on decently priced hardware, and so on. They certainly have their uses, and the field in general has advanced faster than my early expecta…

Yep. These small models are actually worse than GPT 3.5 at some tasks (like recalling facts). You can definitely make models smarter at specific tasks (like tool calling, coding) but you can't compress the entire human knowledge into a 30GB file. It's just not enough bits.

that is why we enable web search for the agent. the memory can come from the internet.

deepseek-v4-flash needs web search to return true facts.

Post reply on HN