Since I like having more than one local LLM to switch between when analysing tricky development issues I decided to try out this new MoE model today. It's a 30B A3B which means it's basically a drop-in replacement for Qwen 3.6 35B A3B with suitable llama.cpp parameters the same.

On their own published benchmark metrics it's supposed to be slightly worse than Qwen, but so far it's not something I've noticed. It's tuned to work well in Opencode which is how I'm running it as well.

Try it out, see how it works for you. I know that there are those who would rather use a Canadian than Chinese model in today's political climate and it does seem to perform better than Gemma 4 at least for me. Just don't forget to use the PR linked from unsloth's description until it has been merged into main.

Maybe it was just me, but in case others have done the same this post might help someone else too.

I have a workstation with plenty of CPU and system RAM, but I'm "GPU poor" in that I only have a 5060Ti with its 16GB of VRAM. Additionally, I need to use the GPU for regular system activities too which means I only have around ~14GB of VRAM available for the LLM.

I'm exclusively using this setup for development and system management tasks, and I've found Qwen 3.6 35B A3B to excel compared to other models. I don't have the VRAM to run the 27GB dense model, so I've spent time on getting the best usage out of the MoE.

Or so I thought. Since "everyone" says to use Unsloth UD-Q4_K_XL that's the quant I've been using, and I've gone a bit back'n'forth with MTP/no MTP, UB increase, mmproj since I've also started using a browser MCP etc.

Today I took another look at their quant chart and thought that since it's MoE maybe I could run Q5_K_S which would be a step up?

Well. Now I'm using Q6_K because it turns out I could run that with the exact same settings as I've optimized my Q4_K_XL setup for which means there are no drawbacks - just a better performing model. I've already noticed how it's able to get out of loops while before I had to interrupt it sometimes.

This is my setup. I get >1000 t/s prefill and >20 t/s inference. I'm not chasing faster inference since I actively read the thought process when working the LLM - but I've increased ub to get faster prefill since that's just waiting time otherwise.

./llama-server
    -hf unsloth/Qwen3.6-35B-A3B-GGUF:UD-Q6_K \
    -c 160000 \
    -n 32768 \
    -fa on \
    -ub 2048 \
    -ctk q8_0 \
    -ctv q8_0 \
    --no-mmap \
    --mlock \
    --no-warmup \
    --chat-template-kwargs '{"preserve_thinking": true}' \
    --temp 0.6 \
    --top-p 0.95 \
    --top-k 20 \
    --min-p 0.0 \
    --presence-penalty 0.0 \
    --repeat-penalty 1.0 \
    --host 0.0.0.0

I also use Opencode with the DCP and Superpowers plugins, which make a tremendous difference both to context handling as well as planning. I have no need for a larger context - I even compact early quite often since the tasks get done before reaching the limit.

Seeing as the game happily lists Mithral and Storm Leather as things that can be used for equipment and upgrades without them existing (currently) in the survival portion of the game, is it the same with the memories?

I'm at 196/200 for the last things to unlock and it's getting very difficult to find something new. The progress bar indicates there would be ~50 more things which seems ... incredible.

Anybody knows?

74% of Ukrainians support fighting Russia even without U.S. assistance. A significant majority—59% of respondents—also believe that Ukraine can defeat Russia on the battlefield

only 6% of respondents said they were willing to make territorial concessions regarding areas occupied by Russia after the full-scale invasion in 2022

Additionally, 70% of respondents are against lowering the mobilization age,

Original article is paywalled, quotes from https://ukrainetoday.org/74-of-ukrainians-ready-to-resist-russia-without-u-s-aid-support-zelenskyys-actions/

Swedish author and famous pro-Ukraine blogger Lars Wilderäng (Cornucopia) reports today that the Swedish security expert Karl Emil Nikka has revealed that Kagi is using the Kremlin propaganda tool Yandex as a backend for searches.

Wilderäng speculates this might mean search terms are leaking to Russia, while others worry about how Kremlin thus can get their talking points into western search results.

Security expert Karl Emil Nikka tells us that the search engine Kagi, popular among tech geeks, uses Russian Yandex, which was introduced after the full-scale invasion. This, of course, gives Russia the opportunity to look at what is searched for via Kagi.

Link (in Swedish), see 11:22 update: https://cornucopia.se/2024/10/uppdateras-ryssland-medger-bruk-av-c-stridsmedel-mot-ukraina-rysk-pilot-som-mordade-68-ukrainare-ihjalslagen-med-hammare-bland-de-allra-storsta-ryska-forlusterna-under-kriget-igar/