this post was submitted on 30 Aug 2026
0 points (NaN% liked)

LocalLLaMA

5238 readers
8 users here now

Welcome to LocalLLaMA! Here we discuss running and developing machine learning models at home. Lets explore cutting edge open source neural network technology together.

Get support from the community! Ask questions, share prompts, discuss benchmarks, get hyped at the latest and greatest model releases! Enjoy talking about our awesome hobby.

As ambassadors of the self-hosting machine learning community, we strive to support each other and share our enthusiasm in a positive constructive way.

Rules:

Rule 1 - No harassment or personal character attacks of community members. I.E no namecalling, no generalizing entire groups of people that make up our community, no baseless personal insults.

Rule 2 - No comparing artificial intelligence/machine learning models to cryptocurrency. I.E no comparing the usefulness of models to that of NFTs, no comparing the resource usage required to train a model is anything close to maintaining a blockchain/ mining for crypto, no implying its just a fad/bubble that will leave people with nothing of value when it burst.

Rule 3 - No comparing artificial intelligence/machine learning to simple text prediction algorithms. I.E statements such as "llms are basically just simple text predictions like what your phone keyboard autocorrect uses, and they're still using the same algorithms since <over 10 years ago>.

Rule 4 - No implying that models are devoid of purpose or potential for enriching peoples lives.

founded 3 years ago
MODERATORS
 

Qwen3.6-27B and Qwen3.8-27B are great. They are fast and quite accurate for not having to spend too much time thinking or searching, as other models with comparable scores have to.

But Qwen3.8-Flash-Next, which tops a lot of benchmarks in its category, is the most obvious Claude distill ever, and although I don't particularly care about how models get trained, I hate the way it sounds and interacts with me.

  • Three warts I'd still fix
  • Statements are down to the two things code can't say
  • The bug is fixed at the substrate that caused it
  • Genuinely good now
  • Also worth a conscious nod
  • Say the word on 1 and/or 3 and I'll do them

Not to mention Claude's over-the-top comments and git commit messages that make you want to scream.

I wish Alibaba would go back to adopting its own style. Qwen3.6-27B and Qwen3.8-27B sound much better.

all 8 comments
sorted by: hot top controversial new old
[–] MagicShel@lemmy.zip 0 points 1 month ago (2 children)

I hate Claude's voice so much. So much ceremony spent on a few simple questions and caveats. I feel like AI defaults to assuming you have no idea what you're doing. ChatGPT is the same but with different verbal tics.

[–] possiblylinux127@lemmy.zip 0 points 1 month ago

Let's cut down to the heart of the matter as you hit the nail squarely on the head

[–] SirDimples@programming.dev 0 points 1 month ago (1 children)

Never used Claude, but GPT-5.6 Sol annoyed me so much I actively avoid it and just use Terra on max. Sol is trained to favor convoluted San Francisco jargon that projects any trivial idea into some sort of an ingenious business maneuver.

[–] fonix232@fedia.io 0 points 1 month ago

I did notice the same. Both Sol and Claude prefer pushing that typical corpo business/product talk that makes me go crazy.

More specifically it keeps talking like that one colleague who speaks fluent corpo BS and manages to amaze non-tech people by verbose statements that ultimately say nothing or are super vague on details.

[–] BeefAndPoultry@lemmus.org 0 points 1 month ago (1 children)

I only use Qwen for Zoo Code, when chatting I just use Gemma

[–] inari@piefed.zip 0 points 1 month ago

That seems to be the consensus on Reddit, Qwen for code, Gemma for everything else. At least it was so for 3.6, not sure if 3.8 changes the situation.

[–] melfie@lemmy.zip 0 points 1 month ago

I have to use Claude at work and have found myself wishing I could use Qwen 3.8 27B at work so I don’t have to waste time deciphering a machine’s fart sniffing. Guess it’s a good thing I don’t have enough VRAM to run a 125B model.