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

391–400 of 563 posts

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

#391

Post by quinnjh »

duzer65657 wrote: post #341
rafaelmn wrote: post #314

Earlier quoted context omitted.

I mean look at the result where he asked about a unicycle - the model couldn't even keep the spokes inside the wheels - would be rudimentary if it "learned" what it means to draw a bicycle wheel and could transfer that to unicycle.

it's the frame that's surprisingly - and consistentnly - wrong. You'd think two triangles would be pretty easy to repro; once you get that the rest is easy. It's not like he's asking "draw a pelican on a four-bar linkage suspension mountainbike..."

is it possible to have greater success with the specificity? I don't think i ever drew a bike frame properly as a kid despite riding them and understanding the concept of spokes and wheels...

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

#392

Post by cpburns2009 »

cpburns2009 wrote: post #325

Anyone else getting gibberish when running unsloth/Qwen3.6-35B-A3B-GGUF:UD-IQ4_XS on CUDA (llama.cpp b8815)? UD-Q4_K_XL is fine, as is Vulkan in general.

Apparently it's a known issue with CUDA 13.2 [1].

[1] https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF/discussi...

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

#393

Post by 999900000999 »

badsectoracula wrote: post #346
999900000999 wrote: post #219

Looking to move off ollama on Open Suse tumbleweed. Should I use brew to install llma.ccp or the zypper to install the tumbleweed package?

You can compile it from source, all you need to do is clone the repository and do a `cmake -B build -DGGML_VULKAN=1` (add other backends if you want) followed by a `cmake --build build --config Release` and then you get all the llama tools in the `build/bin` (including `llama-server` which provides a web-based interface). There is a `docs/build.md` that has more detailed info (especially if you need another backend,…

Cool, I assume this is how adults use llms.

I’m on a nvidia gpu , but I want to be able to combine vram with system memory.

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

#395

Post by withinboredom »

kelnos wrote: post #358
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'm not sure how you can give the flamingo win to Qwen: * It's sitting on the tire, not the seat. * Is that weird white and black thing supposed to be a beak? If so, it's sticking out of the side of its face rather than the center. * The wheel spokes are bizarre. * One of the flamingo's legs doesn't extend to the pedal. * If you look closely at the sunglasses, they're semi-transparent, and the flamingo only has one e…

He literally said it came down to the comment in the SVG. Points for taste, not correctness. Basically.

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

#396

Post by adrian_b »

kylehotchkiss wrote: post #252
zozbot234 wrote: post #40

Earlier quoted context omitted.

This is just one model in the Qwen 3.6 series. They will most likely release the other small sizes (not much sense in keeping them proprietary) and perhaps their 122A10B size also, but the flagship 397A17B size seems to have been excluded.

How many people/hackernews can run a 397b param model at home? Probably like 20-30.

The 397B model can be run at home with the weights stored on an SSD (or on 2 SSDs, for double throughput).

Probably too slow for chat, but usable as a coding assistant.

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

#397

Post by SwellJoe »

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.

Pelicanmaxxing

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

#398

Post by xienze »

adrian_b wrote: post #396
kylehotchkiss wrote: post #252

Earlier quoted context omitted.

How many people/hackernews can run a 397b param model at home? Probably like 20-30.

The 397B model can be run at home with the weights stored on an SSD (or on 2 SSDs, for double throughput). Probably too slow for chat, but usable as a coding assistant.

I think you have that backwards. Agentic coding is way more demanding than simple chat. The request/response loops (tool calling) are much tighter and more numerous, and the context is waaaaay bigger in general.

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

#399

Post by Zopieux »

vlapec wrote: post #362
wrxd wrote: post #210

Earlier quoted context omitted.

Same here. I really hope in a near future local model will be good enough and hardware fast enough to run them to become viable for most use cases

No need to hope; it is inevitable.

Is it inevitable though? Open-weight models large enough to come close to an API model are insanely expensive to run for con/prosumers. I'd put the “expensive” bar at ≥24GB since that's already well into 4 digits, which gives you quite many months of a subscription, not including the power will for >400W continuous.

Color me pessimistic, but this feels like a pipe dream.

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

#400

Post by adrian_b »

JKCalhoun wrote: post #217
WithinReason wrote: post #106

Earlier quoted context omitted.

It's on the page: Precision Quantization Tag File Size 1-bit UD-IQ1_M 10 GB 2-bit UD-IQ2_XXS 10.8 GB UD-Q2_K_XL 12.3 GB 3-bit UD-IQ3_XXS 13.2 GB UD-Q3_K_XL 16.8 GB 4-bit UD-IQ4_XS 17.7 GB UD-Q4_K_XL 22.4 GB 5-bit UD-Q5_K_XL 26.6 GB 16-bit BF16 69.4 GB

"16-bit BF16 69.4 GB" Is that (BF16) a 16-bit float?

The IEEE standard FP16 is an older 16-bit format, which has balanced exponent and significand sizes.

It has been initially supported by GPUs, where it is useful especially for storing the color components of pixels. For geometry data, FP32 is preferred.

In CPUs, some support has been first added in 2012, in Intel Ivy Bridge. Better support is provided in some server CPUs, and since next year also in the desktop AMD Zen 6 and Intel Nova Lake.

BF16 is a format introduced by Google, intended only for AI/ML applications, not for graphics, so initially it was implemented in some of the Intel server CPUs and only later in GPUs. Unlike FP16, which is balanced, BF16 has great dynamic range, but very low precision. This is fine for ML but inappropriate for any other applications.

Nowadays, most LLMs are trained preponderantly using BF16, with a small number of parameters using FP32, for higher precision.

Then from the biggest model that uses BF16, smaller quantized models are derived, which use 8 bits or less per parameter, trading off accuracy for speed.

Post reply on HN