I literally said I’m using qwen3.5:122b for coding. I also use GLM-5 but it’s slightly slower so I generally stick with qwen.

It’s right there, in ollama’s library: https://ollama.com/library/qwen3.5:122b

The weights and everything else for it are on Huggingface: https://huggingface.co/Qwen/Qwen3.5-122B-A10B

This is not speculation. That’s what I’m actually using nearly every day. It’s not as good as Claude Code with Opus 4.6 but it’s about 90% of the way there (if you use it right). When GLM-5 came out that’s when I cancelled my Claude subscription and just stuck with Ollama Cloud.

I can use gpt-oss:20b on my GPU (4060 Ti 16GB)—and it works well—but for $20/month, the ability to use qwen3.5 and GLM-5 are better options.

I still use my GPU for (serious) image generation though. Using ChatGPT (DALL-E) or Gemini (Nano Banana) are OK for one-offs but they’re slow AF compared to FLUX 2 and qwen’s image models running locally. I can give it a prompt and generate 32 images in no time, pick the best one, then iterate from there (using some sophisticated ComfyUI setups). The end result is a superior image than what you’d get from Big AI.

I just added up how much it would cost (in theory—assuming everything is in-stock and ready to ship) to build out a data center capable of training something like qwen3.5:122b from scratch in a few months: $66M. That’s how much it would cost for 128 Nvidia B200 nodes (they have 8 GPUs each), infiniband networking, all-flash storage (SSDs), and 20 racks (the hardware).

If OpenAI went bankrupt, that would result in a glut of such hardware which would flood the market, so the cost would probably drop by 40-60%.

Right now, hardware like that is all being bought up and monopolized by Big AI. This has resulted in prices going up for all these things. In a normal market, it would not cost this much! Furthermore, the reason why Big AI is spending sooooo much fucking money on data centers is because they’re imagining demand. It’s not for training. Not anymore. They’re assuming they’re going to reach AGI any day now and when they do, they’ll need all that hardware to be the world’s “virtual employee” provider.

BTW: Anthropic has a different problem than the others with AGI dreams… Claude (for coding) is in such high demand that their biggest cost is inference. They can’t build out hardware fast enough to meet the demand (inference, specifically). For every dollar they make, they’re spending a dollar to build out infrastructure. Presumably—some day—they’ll actually be able to meet demand with what they’ve got and on that day they’ll basically be printing money. Assuming they can outrun their debts, of course.

I personally love glm-5 and qwen3.5, specifically: https://ollama.com/library/qwen3.5:122b

I’ve used them both for coding and they work really well (way better than you’d think). They’re also perfectly capable of the usual LLM chat stuff (e.g. check my grammar) but all the models (even older, smaller ones) are capable of that stuff these days.

For a treat: Have someone show you using some of these models to search the web! It’s amazing. You don’t see ads, you don’t have to comb through 12 pages of search results, and they read the pages that moment (not cached) to give you summaries of the content. So when you click the link to go to the content you know it’s the thing you were looking for. They’re not using a local index of the Internet, they’re searching on your behalf using whatever search engines you configured. It’s waaaaay better than ChatGPT (which uses Bing behind the scenes whether you like it or not) or Gemini (which uses Google, obviously). The (self-hosted) LLM will literally be running curl for you on Google, DuckDuckGo, Bing, or whatever TF else you want (simultaneously) then reading each of the search results and using your prompt to figure out what the most relevant results are. It’s sooooo nice!

FYI: Ollama.com’s library page is actually a great resource for finding info on all the models that can be self-hosted: https://ollama.com/library

You seem to be unaware that it only takes about four NVIDIA HGX H100 nodes (32 GPUs) to train something like qwen3.5:122b. That model is about as good as ChatGPT was six months to a year ago (for the usual use cases). That would take a long ass time though (over a year) so you’d want probably 50-100 HGX H100s (or lots of the newer, cheaper ARM-based hardware devices).

The weights for qwen3.5:122b are open. That means that if you’ve got the hardware (loads of universities and non-profits have waaaay TF more than 4 HGX H100 nodes) you can continue modern AI development. Everything you need is right there on Huggingface! Deepseek’s stuff is also open I think but I forget. Aside: In my head, I hold the qwen models as “the gold standard” based on many articles I’ve read about them but AI moves so fast, there might be better stuff out on any given day! I haven’t read AI news in like a week so I could be all wrong and qwen3.5 is now sooo obsolete, hehe (that’s how it feels to follow AI news, anyway 🤣).

Even more interesting: qwen3.5:122b isn’t just an LLM. It does visual reasoning (e.g. give it a picture of a plant and ask it to identify it, count the number of screws in an image, estimate distances, etc) as well as the usual LLM stuff. You can read all about it here:

https://ollama.com/library/qwen3.5:122b

…and if you install ollama and spend $20 on ollama.com’s cloud service you can actually try it out without having to own enough GPUs to cover the 245+GB requirement. I highly recommend that service! You can try out all the latest & greatest models on your local PC (or phone!) for any purpose you want for a $20. Whenever a new model is out they usually have it up on their servers within a day or two and it’s fast, too.

FYI: I’ve used ollama cloud to evaluate models for coding (web dev with Python back end) and qwen3.5:122b is fantastic. It’s not as good as Claude Opus 4.6 but it’s close (and cheap) enough that you can just make up for the mistakes with extra instances that check the output with a critical eye (the latest trick in AI-based coding to get good output).

For reference, the University of Texas at Austin has data centers with 4,000 NVIDIA Blackwell (B200/GB200) GPUs, Harvard has 1,144 GPUs, and the University of Cambridge & Bristol (in the UK) has some monstrous mix of Intel and AMD GPUs. All three are perfectly capable of training new models from scratch or using continuing development on existing open-weight models like Deepseek and Qwen.

Generative AI isn’t going anywhere. Furthermore, advancements in that space happen so fast that it’s likely that in a few years we won’t need so many GPUs/VRAM to train models. Especially if ternary models (and similar, like Google’s TurboQuant tech) take off.

I know this is a long comment but I want to point something else out: If OpenAI and Anthropic go bust, that would flood the market with cheap GPUs. It would be a total price collapse and you can bet your ass that clever universities and service providers (like Amazon compute, but 3rd party) would snap those up and bring down prices across the board.

Same places as usual: Academia and open source foundations.

That’s where 99% of all advancements in AI come from. You don’t actually think Big AI is paying as many people to do computer science and mathematics research as all the universities in the world (with computer science programs)?

It’s the same shit as always: Big companies commercialize advancements and discoveries made by scientist and researchers from academia (mostly) and give almost nothing back.

Big AI has partnerships with tons of schools and if it weren’t for that, they wouldn’t be advancing the technology as fast as they are. In fact, the only reason why many of these discoveries are made public at all is because of the agreements with the schools that require the discoveries/papers be published (so their school, professors, researchers, and students can get credit).

Like I was saying before: You don’t need a trillion dollars in data centers to do this stuff. Almost all the GPUs and special chips being used (and preordered, sigh) by Big AI are being used to serve their customers (at great expense). Not for training.

Training used to be expensive but so many advancements have been made this is no longer the case. Instead, most of the resources being used in “AI data centers” (and research) is all about making inference more efficient. That’s the step that comes after you give an AI a prompt.

Training a super modern AI model can be done with a university’s data center or a few hundred thousand to a few million dollars of rented GPUs/compute. It doesn’t even take that long!

Generative AI improves at a ridiculously fast rate. In nearly all the ways you could think of: Training, inference (e.g. figuring out user intent), knowledge, understanding, and weirder, fluffier stuff like “creativity” (the benchmarks of which are dubious, BTW).

These dudes could do pushups without using their arms.

Um… Where would it go? I’ve got about 30 models on my machine right now and I download new ones to try out all the time.

Are you suggesting that they’d all just magically disappear one day‽

Assume all the big AI firms die: Anthropic, OpenAI, Microsoft, Google, and Meta. Poof! They’re gone!

Here would be my reaction: “So anyway… have you tried GLM-7? It’s amazing! Also, there’s a new workflow in ComfyUI I’ve been using that works great to generate…”

Generative AI is here to stay. You don’t need a trillion dollars worth of data centers for progress to continue. That’s just billionaires living in an AGI fantasy land.

Either a lot more tools got a lot better,

That’s what it was. Even the free, open source models are vastly superior to the best of the best from just a year ago.

People got into their heads that AI is shit when it was shit and decided at that moment that it was going to be stuck in that state forever. They forget that AI is just software and software usually gets better over time. Especially open source software which is what all the big AI vendors are building their tools on top of.

We’re still in the infancy of generative AI.

Patents on software shouldn’t exist!

What a childish thing.

It will break down into mostly carbon dioxide and water after a 20-200 years depending on how much UV light hits it (and friction). There won’t be anything for worms to eat!

It was a joke! Dry humor.

The core of the earth is very warm already though.

The second question is silly: “what makes a towel dirt?” Worms, of course!

For the towel to become dirt, you need a nice composting environment and 2+ years. It’ll go best if you turn it from time to time.

The final step is for the worms to eat the leftovers. Then the towel has officially become dirt that’s indistinguishable from the other dirt that surrounds it 👍

Beware horse girls! It’s a trap!

Also, the goddess is a lush. Don’t offer her any booze!

Have to hang em out to dry after surgery.

To be fair, lots of people were crucified at the time. Many with the same name as Jesus and many for the same crimes!

In this case, though… It was merely the pubic execution of a leprechaun.

Is that even necessary? Anyone with a CAD tool can recreate the 3D printed parts from a glance and a few specs.

It’s literally a tube. Which—to be fair—is a “weapon of mass destruction” according to President Bush (the other war criminal president).

You don’t need explosives. It has a spot in the front for a camera. One of the new microcontrollers with AI accelerators can do face recognition extremely quickly. It would be possible to use it as an assassination tool.

Even if you changed nothing about the design, the speed and mass of the thing hitting a person in the face could kill.