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

Qwen3.6-35B-A3B: Agentic coding power, now open to all

qwen.ai

Post reply on HN
Print view View original

501–510 of 563 posts

Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all

#501

Post by jwitthuhn »

3836293648 wrote: post #377
jwitthuhn wrote: post #297

Earlier quoted context omitted.

I've been largely using Qwen3.5-122b at 6 bit quant locally for some c++/go/python dev lately because it is quite capable as long as I can give it pretty specific asks within the codebase and it will produce code that needs minimal massaging to fit into the project. I do have a $20 claude sub I can fall back to for anything qwen struggles with, but with 3.5 I have been very pleased with the results.

How much VRAM do you need for that?

128GB on a mac with unified memory. The model itself takes something like 110 of that and then I have ~16 left over to hold a reasonably sized context and 2 for the OS.

I do have a dedicated machine for it though because I can't run an IDE at the same time as that model.

Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all

#502

Post by mettamage »

kanemcgrath wrote: post #356

I have been using Qwen3.5-35B-A3B a lot in local testing, and it is by far the most capable model that could fit on my machine. I think quantization technology has really upped its game around these models, and there were two quants that blew me away Mudler APEX-I-Quality. then later I tried Byteshape Q3_K_S-3.40bpw Both made claims that seemed too good to be true, but I couldn't find any traces of lobotomization doi…

who do you compare it against qwen3.5 27b?

Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all

#503

Post by coder543 »

naasking wrote: post #453
coder543 wrote: post #183

Earlier quoted context omitted.

Artificial Analysis hasn't posted their independent analysis of Qwen3.6 35B A3B yet, but Alibaba's benchmarks paint it as being on par with Qwen3.5 27B (or better in some cases). Even Qwen3.5 35B A3B benchmarks roughly on par with Haiku 4.5, so Qwen3.6 should be a noticeable step up. https://artificialanalysis.ai/models?models=gpt-oss-120b%2Cg... No, these benchmarks are not perfect, but short of trying it yourself,…

Qwen models commonly get accused of benchmaxxing though. Just something to keep in mind when weighing the standard benchmarks.

Every model release gets accused of that, including the flagship models.

Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all

#504

Post by hansmayer »

jubilanti wrote: post #302
simonw wrote: post #270

I've been running this on my laptop with the Unsloth 20.9GB GGUF in LM Studio: https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF/blob/mai... It drew a better pelican riding a bicycle than Opus 4.7 did! https://simonwillison.net/2026/Apr/16/qwen-beats-opus/

I wonder when pelican riding a bicycle will be useless as an evaluation task. The point was that it was something weird nobody had ever really thought about before, not in the benchmarks or even something a team would run internally. But now I'd bet internally this is one of the new Shirley Cards.

Valid points, but you"d think "superintelligence" would "know" how to draw a pelican on a bike?

Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all

#505

Post by Hugsun »

kanemcgrath wrote: post #356

I have been using Qwen3.5-35B-A3B a lot in local testing, and it is by far the most capable model that could fit on my machine. I think quantization technology has really upped its game around these models, and there were two quants that blew me away Mudler APEX-I-Quality. then later I tried Byteshape Q3_K_S-3.40bpw Both made claims that seemed too good to be true, but I couldn't find any traces of lobotomization doi…

Unfortunately, llama.cpp quantization technology has been stagnant for two years. The main quantization developer left or was kicked out of llama.cpp due to an attribution dispute. He created his own fork ik_llama.cpp where he has made multiple new and better quants.

unsloth and byteshape are just using and highlighting features that have been available the whole time. I am very invested in figuring out a solution to this dispute, or some way to get the new quants upstreamed.

Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all

#506

Post by Hugsun »

zozbot234 wrote: post #383
stratos123 wrote: post #367

Earlier quoted context omitted.

One interesting thing about Qwen3 is that looking at the benchmarks, the 35B-A3B models seem to be only a bit worse than the dense 27B ones. This is very different from Gemma 4, where the 26B-A4B model is much worse on several benchmarks (e.g. Codeforces, HLE) than 31B.

> This is very different from Gemma 4, where the 26B-A4B model is much worse on several benchmarks (e.g. Codeforces, HLE) than 31B. Wouldn't you totally expect that, since 26A4B is lower on both total and active params? The more sensible comparison would pit Qwen 27B against Gemma 31B and Gemma 26A4B against Qwen 35A3B.

They're comparing Qwen's moe vs dense (smaller difference) against Gemma's moe vs dense (bigger difference). Your proposed alternative misses the point.

Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all

#508

Post by zozbot234 »

Hugsun wrote: post #506
zozbot234 wrote: post #383

Earlier quoted context omitted.

> This is very different from Gemma 4, where the 26B-A4B model is much worse on several benchmarks (e.g. Codeforces, HLE) than 31B. Wouldn't you totally expect that, since 26A4B is lower on both total and active params? The more sensible comparison would pit Qwen 27B against Gemma 31B and Gemma 26A4B against Qwen 35A3B.

They're comparing Qwen's moe vs dense (smaller difference) against Gemma's moe vs dense (bigger difference). Your proposed alternative misses the point.

Gemma's dense is bigger than its moe's total parameters. You could totally expect the moe to do terribly by comparison.

Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all

#509

Post by mobiuscog »

bwv848 wrote: post #373
simonw wrote: post #270

I've been running this on my laptop with the Unsloth 20.9GB GGUF in LM Studio: https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF/blob/mai... It drew a better pelican riding a bicycle than Opus 4.7 did! https://simonwillison.net/2026/Apr/16/qwen-beats-opus/

I've been trying the Q4_K_M version, and sometimes it gets stuck in a loop. Gemma 4 doesn’t have this issue.

I'm having the same issues, the more I use it. The repetition penalty doesn't seem to help.

I get some really amusing 'reflective' responses, but I think it needs a bit more cooking. Maybe I'll try another variant.

Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all

#510

Post by naasking »

coder543 wrote: post #503
naasking wrote: post #453

Earlier quoted context omitted.

Qwen models commonly get accused of benchmaxxing though. Just something to keep in mind when weighing the standard benchmarks.

Every model release gets accused of that, including the flagship models.

Less so for Gemma-4 because it falls behind Qwen on benchmarks. Tests for benchmaxxing are also strongly suggestive: https://x.com/bnjmn_marie/status/2041540879165403527
Post reply on HN