Are we going to get to the point where a local model can do almost what sonnet 4.6 can do?
Qwen3.6-35B-A3B: Agentic coding power, now open to all
331–340 of 563 posts
Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all
#332My last laptop was a used 2012 T530. My current is a used M1 MBP Pro with 16GB of ram. I thought this was all I was ever going to need, but wanting to run really nice models locally has me thinking about upgrading. Although, part of me wants to see how far I could get with my trusty laptop.
Your current laptop is still a fine thin client. Unless you program in the woods, it's probably cheapest to build a home inference box and route it over Tailscale or something.
Or just an API server for all other devices to connect and do stuff with it.
Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all
#333Earlier 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.
I'm running it on dual DGX Sparks.
Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all
#334Does anyone have any experience with Qwen or any non-Western LLMs? It's hard to get a feel out there with all the doomerists and grifters shouting. Only thing I need is reasonable promise that my data won't be used for training or at least some of it won't. Being able to export conversations in bulk would be helpful.
Personally, I wouldn't trust any foreign or domestic LLM providers to not train on your data. I also wouldn't trust them to not have a data breach eventually which is worse. If you're really worried about your data, run it locally. The Chinese models (Qwen, GLM, etc.) are really competitive to my understanding.
Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all
#335Fingers crossed for mid and larger models as well. I'd personally love to see Qwen3.6-122B-A10B.
That would be really great. Though 3.5 122B is already doing a lot of work in our setup.
Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all
#336Earlier quoted context omitted.
Please publish sha256sums of the merged GGUFs in the model descriptions. Otherwise it's hard to tell if the version we have is the latest.
Why do you merge the GGUFs? The 50 GB files are more manageable (IMO) and you can verify checksums as you say.
I admit it's a habit that's probably weeks out of date. Earlier engines barfed on split GGUFs, but support is a lot better now. Frontends didn't always infer the model name correctly from the first chunk's filename, but once llama.cpp added the models.ini feature, that objection went away.
The purist in me feels the 50GB chunks are a temporary artifact of Hugging Face's uploading requirements, and the authoritative model file should be the merged one. I am unable to articulate any practical reason why this matters.
Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all
#337Is this the hybrid variant of Gwent and Quen? I hope this is in The Witcher IV!
Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all
#338Are we going to get to the point where a local model can do almost what sonnet 4.6 can do?
Of course we are. And Opus 4.6+. It's a matter of when, not if.
Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all
#339I'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/
Interesting, qwen has the pelican driving on the left lane. Coincidence or has it something to do with the workers providing the RL data?
Could be on a bike path where bikes are on the left and pedestrians to the right.
Re: Qwen3.6-35B-A3B: Agentic coding power, now open to all
#340I'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.
They’ll hardcode it in 4.8, just like they do when they need to “fix” other issues