What @mierdabird@lemmy.dbzer0.com said, but the adapters arent cheap. You’re going to end up spending more than the 1060 is worth.

A used desktop to slap it in, that you turn on as needed, might make sense? Doubly so if you can find one with an RTX 3060, which would open up 32B models with TabbyAPI instead of ollama. Some configure them to wake on LAN and boot an LLM server.

You can still use the IGP, which might be faster in some cases.

Oh actually that’s a great card for LLM serving!

Use the llama.cpp server from source, it has better support for Pascal cards than anything else:

https://github.com/ggml-org/llama.cpp/blob/master/docs/multimodal.md

Gemma 3 is a hair too big (like 17-18GB), so I’d start with InternVL 14B Q5K XL: https://huggingface.co/unsloth/InternVL3-14B-Instruct-GGUF

Or Mixtral 24B IQ4_XS for more ‘text’ intelligence than vision: https://huggingface.co/unsloth/Mistral-Small-3.2-24B-Instruct-2506-GGUF

I’m a bit ‘behind’ on the vision model scene, so I can look around more if they don’t feel sufficient, or walk you through setting up the llama.cpp server. Basically it provides an endpoint which you can hit with the same API as ChatGPT.

1650

You mean GPU? Yeah, it’s good, I was strictly talking about purchasing a laptop for LLM usage, as most are less than ideal for the money. Laptop vram pools are relatively small and SO-DIMMS are usually very slow.

Things will get much better once the “Max” AMD SKUs proliferate.

Yeah, just paying for LLM APIs is dirt cheap, and they (supposedly) don’t scrape data. Again I’d recommend Openrouter and Cerebras! And you get your pick of models to try from them.

Even a framework 16 is not good for LLMs TBH. The Framework desktop is (as it uses a special AMD chip), but it’s very expensive. Honestly the whole hardware market is so screwed up, hence most ‘local LLM enthusiasts’ buy a used RTX 3090 and stick them in desktops or servers, as no one wants to produce something affordable apparently :/

I was a bit mistaken, these are the models you should consider:

https://huggingface.co/mlx-community/Qwen3-4B-4bit-DWQ

https://huggingface.co/AnteriorAI/gemma-3-4b-it-qat-q4_0-gguf

https://huggingface.co/unsloth/Jan-nano-GGUF (specifically the UD-Q4 or UD-Q5 file)

they are state-of-the-art at this size, as far as I know.

8GB?

You might be able to run Qwen3 4B: https://huggingface.co/mlx-community/Qwen3-4B-4bit-DWQ/tree/main

But honestly you don’t have enough RAM to spare, and even a small model might bog things down. I’d run Open Web UI or LM Studio with a free LLM API, like Gemini Flash, or pay a few bucks for something off openrouter. Or maybe Cerebras API.

…Unfortunely, LLMs are very RAM intensive, and >4GB (more realistically like 2GB) is not going to be a good experience :(

Actually, to go ahead and answer, the “fastest” path would be LM Studio (which supports MLX quants natively and is not time intensive to install), and a DWQ quantization (which is a newer, higher quality variant of MLX models).

Hopefully one of these models, depending on how much RAM you have:

https://huggingface.co/mlx-community/Qwen3-14B-4bit-DWQ-053125

https://huggingface.co/mlx-community/Magistral-Small-2506-4bit-DWQ

https://huggingface.co/mlx-community/Qwen3-30B-A3B-4bit-DWQ-0508

https://huggingface.co/mlx-community/GLM-4-32B-0414-4bit-DWQ

With a bit more time invested, you could try to set up Open Web UI as an alterantive interface (which has its own built in web search like Gemini): https://openwebui.com/

And then use LM Studio (or some other MLX backend, or even free online API models) as the ‘engine’

Alternatively, especially if you have a small RAM pool, Gemma 12B QAT Q4_0 is quite good, and you can run it with LM Studio or anything else that supports a GGUF. Not sure about 12B-ish thinking models off the top of my head, I’d have to look around.

Honestly perplexity, the online service, is pretty good.

As for local running, one question first: how much RAM does your Mac have? This is basically the factor for what model you can and should run.

I don’t understand.

Ollama is not actually docker, right? It’s running the same llama.cpp engine, it’s just embedded inside the wrapper app, not containerized. It has a docker preset you can use, yeah.

And basically every LLM project ships a docker container. I know for a fact llama.cpp, TabbyAPI, Aphrodite, Lemonade, vllm and sglang do. It’s basically standard. There’s all sorts of wrappers around them too.

You are 100% right about security though, in fact there’s a huge concern with compromised Python packages. This one almost got me: https://pytorch.org/blog/compromised-nightly-dependency/

This is actually a huge advantage for llama.cpp, as it’s free of python and external dependencies by design. This is very unlike ComfyUI which pulls in a gazillian external repos. Theoretically the main llama.cpp git could be compromised, but it’s a single, very well monitored point of failure there, and literally every “outside” architecture and feature is implemented from scratch, making it harder to sneak stuff in.

OK.

Then LM Studio. With Qwen3 30B IQ4_XS, low temperature MinP sampling.

That’s what I’m trying to say though, there is no one click solution, that’s kind of a lie. LLMs work a bajillion times better with just a little personal configuration. They are not magic boxes, they are specialized tools.

Random example: on a Mac? Grab an MLX distillation, it’ll be way faster and better.

Nvidia gaming PC? TabbyAPI with an exl3. Small GPU laptop? ik_llama.cpp APU? Lemonade. Raspberry Pi? That’s important to know!

What do you ask it to do? Set timers? Look at pictures? Cooking recipes? Search the web? Look at documents? Do you need stuff faster or accurate?

This is one reason why ollama is so suboptimal, with the other being just bad defaults (Q4_0 quants, 2048 context, no imatrix or anything outside GGUF, bad sampling last I checked, chat template errors, bugs with certain models, I can go on). A lot of people just try “ollama run” I guess, then assume local LLMs are bad when it doesn’t work right.

Totally depends on your hardware, and what you tend to ask it. What are you running? What do you use it for? Do you prefer speed over accuracy?

15M views

TBH you should fold this into localllama? Or open source AI?

I have very mixed (mostly bad) feelings on ollama. In a nutshell, they’re kinda Twitter attention grabbers that give zero credit/contribution to the underlying framework (llama.cpp). And that’s just the tip of the iceberg, they’ve made lots of controversial moves, and it seems like they’re headed for commercial enshittification.

They’re… slimy.

They like to pretend they’re the only way to run local LLMs and blot out any other discussion, which is why I feel kinda bad about a dedicated ollama community.

It’s also a highly suboptimal way for most people to run LLMs, especially if you’re willing to tweak.

I would always recommend Kobold.cpp, tabbyAPI, ik_llama.cpp, Aphrodite, LM Studio, the llama.cpp server, sglang, the AMD lemonade server, any number of backends over them. Literally anything but ollama.

…TL;DR I don’t the the idea of focusing on ollama at the expense of other backends. Running LLMs locally should be the community, not ollama specifically.

At risk of getting more technical, some near-future combination of bitnet-like ternary models, less-autoregressive architectures, taking advantage of sparsity, and models not being so stupidly general-purpose will bring inference costs down dramatically. Like, a watt or two on your phone dramatically. AI energy cost is a meme perpetuated by Altman so people will give him money, kinda like a NFT scheme.

…In other words, it’s really not that big a deal. Like, a drop in the bucket compared to global metal production or something.

The cost of training a model in the first place is more complex (and really wasteful at some ‘money is no object’ outfits like OpenAI or X), but also potentially very cheap. As examples, Deepseek and Flux were trained with very little electricity. So was Cerebras’s example model.

Yyep. That’s a win.

The comic has basically been me for the last decade.

Laughs uncomfortably.

It’s politicized.

It even works in hindsight. I pointed out some cherished fan remaster of a TV show made years ago was machine learning processed, which apparently everyone forgot. I got banned from the fandom subreddit for the no AI rule.

The ironic thing is this works in corpo AI slop’s favor, as anti-AI sentiment hurt locally runnable, open weight models and earnest efforts more than anything.

Let’s look at a “worst case” on my PC. Let’s say 3 attempts, 1 main step, 3 controlnet/postprocessing steps, so 64-ish seconds of generation at 300W above idle.

…That’s 5 watt hours. You know, basically the same as using photoshop for a bit. Or gaming for 2 minutes on a laptop.

Datacenters are much more efficient because they batch the heck out of jobs. 60 seconds on a 700W H100 or MI300X is serving many, many generations in parallel.

Not trying to be critical or anything, I hate enshittified corpo AI, but that’s more-or-less what generation looks like.

Apparently so.

Slop means shitty and disingenuous to me, which is not this comic. Frankly, it hits me hard.

There’s a word for that:

A sausage fest.

And not the fun kind.