- 9 Posts
- 29 Comments
3·3 days agoThey don’t tell us things lol. All they said was
Which sounds like there’s something “better” coming, but doesn’t deny the possibility of 35b-a3b. Which is weird because “better” is subjective and depends on your hardware. It could be smaller and smarter than 3.6 35b, but then people are gonna ask for a 3.8 35b because it should be even smarter.
They have a graph, differences are tiny at that high end
I feel like Ornith 1.0 9b was the best coding model at that size (since Qwen has neglected that size). Maybe now Ling Tiny is better but I’m curious to try Ornith 1.5
3·6 days agoI would suggest using llama.cpp instead of Ollama, or maybe Unsloth Studio
Sounds like you could just replace the template to fix it, there’s a popular Qwen fixed template on hugging face, try that
https://huggingface.co/froggeric/Qwen-Fixed-Chat-Templates/blob/main/chat_template.jinja
1·9 days agollama.cpp is way better than Ollama. If you want a GUI then use Unsloth Desktop
Use llama.cpp or Unsloth Desktop with an uncensored/abliterated model. Gemma 4 models are great for chats. But I don’t think it would give you better answers than the people in this community. And I would be too afraid to give it a web search MCP while asking it questions about piracy.
A better place to ask about AI stuff would be !localllama@sh.itjust.works
I guess botched remasters aside lol, there isn’t really a limit to how poorly something can be done
There are other ways investigators could have connected activity to Ono. He may have used Nyaa without a VPN while logged into an account. Data from website cookies or content-delivery-network logs could also have been matched with Nyaa activity and tracker records. CODA has not said whether any of those methods were used.
VPN isn’t enough, you gotta worry about fingerprinting and cookies too. Recently it was revealed that Windows 11 has strong fingerprinting options that could be used to track you.
I wonder if they were using a VPN or seedbox. You could even do remote desktop into seedbox and browse the sites from there.
Old music doesn’t get re-recorded for digital releases, they pull it from the original master again, which is better and more authentic than vinyl.
I mean if you went to a concert would you want them to add hiss and pop to the performance so it sounds more like vinyl? Lol certainly you wouldn’t want them to intentionally worsen the signal-to-noise ratio
my laptop is crappy, so like 5 tokens per second lol, prompt processing of like 20 tokens per second
I think a decent laptop nowadays, even running CPU only, could probably do like 5x faster
I’ve run Qwen 3.5 4b and Gemma 4 e2b on CPU only, this should be faster than those I think (fewer active parameters). If you have AVX512 or AVX10 then it should help a bit. Still slow compared to a GPU lol.
anyone try this? this might be good for my crappy laptop lol
is it good enough to use with Zoo Code? is it better than Qwen 3.5 4b?
EDIT: woa
https://artificialanalysis.ai/models/ling-3-0-tiny
But not yet supported in llama.cpp https://github.com/ggml-org/llama.cpp/pull/26608
Actually funny he’s not asking it to work harder (that would be system prompt or user message), he’s forcing it to think that it will work harder
That’s a really cool idea. It’s like inception for an LLM, you make it think it was the one that thought of this lol
Have you tried preserve thinking? https://lemmus.org/post/24365786
(Oops I got my Gemma and Qwen speeds mixed up, edited the post to fix it.)
But now with the new commits they added, with the same number of hot experts, Qwen is up to about 34. If I increase hot experts to 48 then I get around 37.
Gemma is still around 23 with just 10 hot experts. With 16 hot experts I get about 26 TPS. If I overprovision my VRAM (thanks to
GGML_CUDA_ENABLE_UNIFIED_MEMORY=1) then 24 hot experts can give me 29 TPS, and 32 hot experts 34 TPS.
In a few minutes a significant performance improvement incoming
👀
Yes this can run Qwen 3.6 35b-a3b pretty nicely! And they might be releasing an updated version of that soon. Your BC-250 only has 16GB total which is not enough for 35b.
I also have 32GB RAM and 8GB VRAM, my computer is a little slower than yours, see my guide: https://lemmus.org/post/24235317
For the BC-250 you might try smaller models like Ling 3.0 Tiny, Ornith 1.5 9b, or Gemma 4 12b QAT
For your 8GB RAM devices, you can run Gemma 4 e4b QAT, Qwen 3.5 4b, or maybe Ling 3.0 Tiny
Use Unsloth Desktop or llama.cpp. Then you can connect Zoo Code to it, that’s a VSCode extension which I like for programming with my local LLMs.