My experience with local LLM

When I first got into local LLMs nearly 3 years ago, in mid 2023, the frontier closed models were ofcourse impressively capable.

I then tried my hand on running 7b size local models, primarily one called Zephyr-7b (what happened to these models?? Dolphin anyone??), on my gaming PC with 8GB AMD RX580 GPU. Fair to say it was just a curiosity exercise (in terms of model performance).

Fast forward to this month, I revisit local LLM. (Although I no longer have the gaming PC, cost-of-living-crisis anyone 😫 )

And, the 31b size models look very sufficient. #Qwen has taken the helm in this order. Which is still very expensive to setup locally, although within grasp.

I'm rooting for the edge-computing models now - the ~2b size models. Due to their low footprint, they are practical to run in a SBC 24/7 at home for many people.

But these edge models are the 'curiosity category' now.

48 points · 18 comments · view on lemmy.world

18 Comments

SuspciousCarrot78@lemmy.world · 13 pts · 139d
[ removed ]
umbrella@lemmy.ml · 11 pts · 139d (11 replies)
[ removed ]
SuspciousCarrot78@lemmy.world · 14 pts · 138d (6 replies)
[ removed ]
umbrella@lemmy.ml · 1 pts · 138d (5 replies)
[ removed ]
SuspciousCarrot78@lemmy.world · 5 pts · 137d (1 reply)
[ removed ]
umbrella@lemmy.ml · 0 pts · 135d
[ removed ]
ntn888@lemmy.ml · 3 pts · 138d (2 replies)

I didn;t try any 7b ones lately, they may be better fit for 16gb I think. I was able to try the 2b ones as I mentioned (on cpu). they are subpar. like mentioned the usable ones were 31b, I think you need atleast 24gb vram for most models though. maybe someone else can suggest better.

umbrella@lemmy.ml · 1 pts · 138d (1 reply)
[ removed ]
ntn888@lemmy.ml · 2 pts · 138d

you can give "unloading some layers to RAM" a try though.. that way you can get your hands on the "usable" 31b models. browse around to find some good 31b ones.. GL

Samskara@sh.itjust.works · 0 pts · 135d (3 replies)

Do you have 24 GB?

umbrella@lemmy.ml · 2 pts · 135d (2 replies)
[ removed ]
Samskara@sh.itjust.works · 0 pts · 135d (1 reply)

That’s your issue.

umbrella@lemmy.ml · 1 pts · 135d
[ removed ]
inconel@lemmy.ca · 7 pts · 138d (3 replies)

For small model bonsai series seems getting the spotlight. Natively trained on1bit and ternary 1.58bit, 8B runs on ~1GB memory. I'm curios on local models but haven't tried because of lack of gaming rig but it seems work enough for regular pc

SuspciousCarrot78@lemmy.world · 4 pts · 138d
[ removed ]
ntn888@lemmy.ml · 1 pts · 138d

funny I tried the 8B bonsai https://huggingface.co/prism-ml/Bonsai-8B-gguf when loaded it takes ~7GB RAM!! When prompting it stalls my llama.cpp container (I'm running on a weak 4th gen i5)

ntn888@lemmy.ml · 1 pts · 138d

Interesting thanks!

PixelatedSaturn@lemmy.world · 7 pts · 139d (7 replies)

For what stuff do you want to use them? I don't think they come remotely close to today's commercial models. Maybe for a specific purpose?

ntn888@lemmy.ml · 6 pts · 139d (6 replies)

hey, thanks for your response.. yeah that's what I meant, the 2b models aren't usable in today's state, but more practical for everyday use if they work out..

I actually meant the 31b models are useful for my purpose. I don't do full-on agentic coding, just interactive chat/prompting. Example, I make good use for making linux shell scripts (as I don't know howto myself). Currently I use qwen3.5-flash via cloud. It's as good as the frontier models back then if not better..

PixelatedSaturn@lemmy.world · 3 pts · 139d (4 replies)

I wanted to use smaller models, but then do more work on the "thinking" process. I didn't come far, because it get so slow with normal hardware and too expensive on dedicated one. Time consuming (I'm also not a programmer) but a fun project, but in the end I just decided to satisfy the privacy angle with protons ai Lumo.

inari@piefed.zip · 1 pts · 139d (3 replies)

Proton has AI? Damn, that's gotta be bleeding their coffers

SuspciousCarrot78@lemmy.world · 7 pts · 139d (2 replies)
[ removed ]
PixelatedSaturn@lemmy.world · 2 pts · 138d (1 reply)

They have been working on this. Only 3 months ago it was pretty terrible. Today it's almost on par with chatgpt. A bit worse on rag, slower,.. good enough for normal use.

SuspciousCarrot78@lemmy.world · 1 pts · 138d
[ removed ]
SuspciousCarrot78@lemmy.world · 3 pts · 139d
[ removed ]
ZoteTheMighty@lemmy.zip · 6 pts · 138d (4 replies)

This weekend I had an LLM walk me through setting up some home server stuff and networking. I tried using Proton's Lumo and Qwen 3.6 locally. I have to say Qwen was the more impressive of the two models. When I first tried running models locally like llama 4, I remember thinking to myself that this was a dead end and big servers would always have the advantage, but it seems like we're hitting a turning point where many things can be done locally.

ntn888@lemmy.ml · 1 pts · 138d (3 replies)

cool what was your hardware, and which qwen size you used? thanks

ZoteTheMighty@lemmy.zip · 3 pts · 137d (2 replies)

I have a 24GB AMD 7900XTX, and it's a 35b parameter model.

ericwdhs@discuss.online · 4 pts · 137d

Ooo... I'm running a 7900 XTX as well. Having 24GB without the Nvidia tax has been super nice for AI stuff. I have a 16GB 6900 XT running in another computer, and a lot of my AI model selection is still sized for it. I may need to stop procrastinating and copy your setup sooner rather than later.

ericwdhs@discuss.online · 1 pts · 137d

Before I forget, can I ask you what GPU driver version you're running? I recently encountered some stability issues after a driver update (trying to support gaming and AI stuff at the same time), and the latest version I could find any stability claims for was 24.12.1.

fozid@feddit.uk · 4 pts · 138d

For me, anything less than gpt oss 20b (a2b) is just for messing around with or for basic categorisation and basic text or data processing with highly structured prompts.