lol basically

Sure, a benchmark doesn’t capture all the subtleties and different use cases, but it does give a general idea of the capabilities of a model. Obviously, you have to run the model and see if it does what you need. But the chart isn’t really about the nuance, it’s showing how drastically the efficiency of the models has improved in just a year. The fact that we can even reasonably compare a model you can run on a desktop to one that needed a data center just a year ago is phenomenal.

ah gotcha, and I’ve made the list myself apparently

I’m fairly optimistic that people will figure out how to optimize the models a lot further going forward. One obvious path is to try and separate the reasoning network from the trivia that gets baked into the model, and some work is being done in this area. If you could have a context free reasoning engine and then feed the facts it needs to know on the fly based on the context you’re running it in, then you could likely have a much smaller model that’s very capable.

Not sure what Moore’s law has to do with anything here to be honest. The models you can run locally on a consumer desktop can do real work, and their resource usage is no different from any other software like games that you’d run.

You need a GPU with around 16gb vram at a minimum to run qunatized version.

that’s the other huge advantage of open models you can run locally

The difference is that you can run Qwen completely local though.

yeah, it’s not a completely insane amount of data, and a db like postgres can do fast text search on that too with fuzzy matching

Ok, but that’s a completely nonsensical statement. If you ever used Qwen in an agentic loop, you’d know that it delivers working code, and it takes about same resources as playing a modern game, and I don’t see anybody whinging that game are too inefficient for what they deliver.

I mean baking knowledge into a model isn’t really all that useful to begin with. Just download wikipedia locally and have it access it through tool use, it’s way more efficient and more accurate. And yeah, I find Q6 tends to be the sweet spot where it’s close enough to full 16 bit in performance, but doesn’t chew up too much memory.

some turbolib made a Lemmy client that had a hardcoded blacklist of users and instances they considered to be communists

https://lemmy.ml/post/50384766/26835474

Ok that’s fair, what Iran is doing is very similar to Russia’s approach of launching combination of drones and missiles to overwhelm Ukrainian defences.

and it’s even lower with MTP https://huggingface.co/unsloth/Qwen3.6-27B-MTP-GGUF

lol yeah, dying gasps for the empire here

I don’t really see the misleading part in the headline myself. While China and Russia undoubtedly help Iran, it seems that Iranians are perfectly capable to inflict massive damage on the empire using their own technical capabilities. While Ukraine is a western proxy entirely dependent on its western masters, Iran has its own deep technological and industrial base. It’s worth remembering that it was Iran that originally did a technology transfer of their drone technology to Russia rather than the other way around.

the system is working as intended

I think if you read the actual article you’ll see that the headline is justified. It explains the actual mechanics of what Iran did, and this shouldn’t be trivialized. While the idea of what to do might not be that complicated, actually implementing it against an adversary like the US is not trivial.

What actually matters here is that the whole US military doctrine has become obsolete because of drones. It’s not just that Iran can find and hit US targets, it’s that they can do it very cheaply while US has to spend expensive and difficult to build interceptors which can’t even hit these kinds of targets effectively. It’s the economics of the whole thing that are defeating the US.

There is no way around needing helium, but China will boost their own production in time and they have access to helium from Russia as well. Meanwhile, the reason they can move so fast is due to having a centrally planned system the west rejects. It’s the same reason Russia is now outproducing all of NATO in terms of weapons and ammunition.

The west is in no position to recreate the necessary supply chains. Rare earths alone is a perfect example. I wrote about the problem in detail here if you’re interested. Even under the most aggressive diversification scenarios, China is expected to retain 80% of global refining capacity by 2040. And that’s just one example. The west has very little leverage at this point.

Also, as she correctly points out, China doesn’t need to get to ASML levels to make these fabs useful. As long as they’re better than the current ones, that already means China will be producing better domestic chips. And they’re also innovating with architecture as well such as Huawei’s tau law design which folds the chip and dramatically improves performance. All these things work together.

Finally, China is aggressively pursuing alternative computing substrates and some are already starting to come online. This is a recent example. These kinds of things have the potential to leapfrog silicon chips entirely. There’s nothing comparable happening in the west because you can’t do these kinds of projects without massive amount of state funding and direction to sustain them until they become profitable on their own.