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

371–380 of 563 posts

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

#371

Post by monksy »

simonw wrote: post #301
cyclopeanutopia wrote: post #290

Earlier quoted context omitted.

But that you also gave a win to Qwen on flamingo is pretty outrageous! :) Tthe right one looks much better, plus adding sunglasses without prompting is not that great. Hopefully it won't add some backdoor to the generated code without asking. ;)

I love how the Chinese models often have an unprompted predilection to add flair. GLM-5.1 added a sparkling earring to a north Virginia opossum the other day and I was delighted: https://simonwillison.net/2026/Apr/7/glm-51/

You're running 5.1 locally or hosted?

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

#372

Post by monksy »

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/

Hey I really enjoy your blog. On some things I end up finding a blog post of yours thats a year+ old and at other times, you and I are investigating similar things. I just pulled Qwen3.6 - 35b -A3B (Can't believe thats a A3B coming from 35b).

I'm impressed about the reach of your blog, and I'm hoping to get into blogging similar things. I currently have a lot on my backlog to blog about.

In short, keep up the good work with an interesting blog!

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

#373

Post by bwv848 »

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.

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

#375

Post by storus »

spuz wrote: post #344
storus wrote: post #251

Earlier quoted context omitted.

You should. 3.5 MoE was worse than 3.5 dense, so expecting 3.6 MoE to be superior than 3.5 dense is questionable, one could argue that 3.6 dense (not yet released) to be superior than 3.5 dense.

Ok but you made a claim about the new model by stating a fact about the old model. It's easy to see how you appeared to be talking about different things. As for the claim, Qwen do indeed say that their new 3.6 MoE model is on a par with the old 3.5 dense model: > Despite its efficiency, Qwen3.6-35B-A3B delivers outstanding agentic coding performance, surpassing its predecessor Qwen3.5-35B-A3B by a wide margin and ri…

This says a slightly different thing:

https://x.com/alibaba_qwen/status/2044768734234243427?s=48&t...

If you look, at many benchmarks the old dense model is still ahead but in couple benchmarks the new 35B demolishes the old 27B. "rivaling" so YMMV.

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

#376

Post by stratos123 »

zengid wrote: post #351

any tips for running it locally within an agent harness? maybe using pi or opencode?

It pretty much just works. Run the unsloth quant in llama.cpp and hook it up to pi. A bunch of minor annoyances like not having support for thinking effort. It also defaults to "interleaved thinking" (thinking blocks get stripped from context), set `"chat_template_kwargs": {"preserve_thinking": True},` if you interrupt the model often and don't want it to forget what it was thinking.

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

#377

Post by 3836293648 »

jwitthuhn wrote: post #297
aliljet wrote: post #73

I'm broadly curious how people are using these local models. Literally, how are they attaching harnesses to this and finding more value than just renting tokens from Anthropic of OpenAI?

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?

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

#378

Post by rubiquity »

zozbot234 wrote: post #366
rubiquity wrote: post #364

Earlier quoted context omitted.

Have you tried running llama.cpp with Unified Memory Access[1] so your iGPU can seamlessly grab some of the RAM? The environment variable is prefixed with CUDA but this is not CUDA specific. It made a pretty significant difference (> 40% tg/s) on my Ryzen 7840U laptop. 1 - https://github.com/ggml-org/llama.cpp/blob/master/docs/build...

Your link seems to be describing a runtime environment variable, it doesn't need a separate build from source. I'm not sure though (1) why this info is in build.md which should be specific to the building process, rather than some separate documentation; and (2) if this really isn't CUDA-specific, why the canonical GGML variable name isn't GGML_ENABLE_UNIFIED_MEMORY , with the _CUDA_ variant treated as a legacy alias…

You are right that it is an environment variable, and that's how I have it set in my nix config. Thanks for correcting that.

Unfortunately llama.cpp is somewhat notorious for having lackluster docs. Most of the CLI tools don't even tell you what they are for.

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

#380

Post by 3836293648 »

sliken wrote: post #214
ru552 wrote: post #81

Earlier quoted context omitted.

You won't like it, but the answer is Apple. The reason is the unified memory. The GPU can access all 32gb, 64gb, 128gb, 256gb, etc. of RAM. An easy way (napkin math) to know if you can run a model based on it's parameter size is to consider the parameter size as GB that need to fit in GPU RAM. 35B model needs atleast 35gb of GPU RAM. This is a very simplified way of looking at it and YES, someone is going to say you…

> You won't like it, but the answer is Apple. Or strix halo. Seems rather over simplified. The different levels of quants, for Qwen3.6 it's 10GB to 38.5GB. Qwen supports a context length of 262,144 natively, but can be extended to 1,010,000 and of course the context length can always be shortened. Just use one of the calculators and you'll get much more useful number.

What Strix Halo system has unified memory? A quick google says it's just a static vram allocation in ram, not that CPU and GPU can actively share memory at runtime
Post reply on HN