kzrdude

I guess Nebula is better for video watching, but without recommendations and video comments, it feels dead. I did pay for it for a year and then gave up. Is the problem me?

I guess it was obscure before, but someone did "know" it: https://old.reddit.com/r/Qwen_AI/comments/1u7xnvq/how_big_is...

Exactly, DeepSeek, Qwen etc are catching the attention because they put out their tech docs and papers, so we can read about how the models work and what they think their innovation was this time.

If I read correctly, that's based on a 1-bit quantization, and can we really expect that to produce any useful output at all?

Not clear to me how it selects that, but it considers it to be #1 out of 173 models "of the same class".

125B regular params, 51B engrams, 4B MTP. Something like that. It should have a label of effectively 125B params with A6B (6B active).

They are giving pointers to how to run it now using for example https://recipes.vllm.ai/Qwen/Qwen3.8-Flash-Next (and an especially provided vllm release).

(Edited: I thought Qwen3.8 Flash Next was smaller, but it's not, in bytes. Here's how they compare.)

DSV4 Flash 304B params, 167 GB download (at full size)

Qwen3.8 Flash Next 180B params, 360 GB download (at full size)

Anyone who likes economics or money should see it for what it is, a tax on every transaction. Of course you don't want anyone to lodge themselves into your market and get 3% skimmed off on every transaction. This, for the good of the free market, the fees must be limited to sub 1%, preferably zero.

I guess this is why English Wikipedia was down. It's back up now.

That reads to me as if the GP confabulated a different exploit target than what the article mentioned (or the article changed), but it's also a threat that should be taken seriously.

Based on your new information, I think it would be good to test the effect of supplemenths vs supplemints.

I agree. But models in difference to compiled binaries, are useful as just weights and can be further refined and post-trained, at least.

I don't know LLM theory well enough to say if there's some secret sauce they can hold back that makes training ineffective. Less effective I'm sure, we don't have access to their smart training schemes, but post-training should always be possible IIUC.

I'm unnerved by how alphazero is more complicated than the "intelligent llms"; it has at least multiple heads and MCTS, a search algorithm. The LLMs seem to just be monolithic (if complicated) architectures where tokens go in the bottom and tokens are spit out at the top.

This is also the problem, we are continuing with LLMs because they fit hardware well, instead of experimenting with trying to model knowledge (how do I know what I know) and metacognition (what do I know).

Current architecture can't tell the difference between fact and fiction, so to speak. (Which I think is a better way to say it rather than truth and falsehood; facts have provenance and context, not just "true" or "false".)

In the feedback loop when working on anything UI or graphical output related.

It seems like attention methods are still evolving, just like residual connections. I’m sure there is a lot left to discover too.

It beats a frontier model from 12 months according to this bench: https://news.ycombinator.com/item?id=49334544

It is not the whole story, and knowledge is very lacking, but it has gotten a lot of attention. That model together with DeepSeek V4 Flash are the highlights of this summer on the open/local models side.

Can it replace pi? Would be good to have a robust and fast replacement.

Maybe ”agent shell” could be a better term for harness. But it’s pretty overloaded..

The former

27B means it has 27 billion parameters, but how that translates to model size depends on the model architecture. For Qwen 3.5 architecture (same as 3.8) the native size is around 55 GB memory. So you need quantization to fit it in 32 GB.

Go here: https://huggingface.co/unsloth/Qwen3.8-27B-GGUF

See the right hand side panel, you see a whole palette of quantizations and their respective sizes. Should give you an idea. Note that these are not the only quantizations available.

Funny but shows how it doesn't have a useful world model

Are they forgetting that people use Claude for presentations and business reports too?

Why is it connected to number of parameters?

This comment by the Z AI lead is relevant, about parameter vs data scaling: https://news.ycombinator.com/item?id=49357405

Mirror: https://xcancel.com/jietang/status/2089941544581403107

Is also a Swedish word, though slang, the imperative form of paxa which basically means the same as "dibs!" If I know my english colloquialisms correctly.

A lot of them are Scandi place names, too.

The German vowels are a good start, but Swedish pronunciation is of course a little bit vowel shifted from German.