This is the best thing I've ever designed. I am so happy with how it turned out, I will be using my b70s Omarchy turns your pc into AI Free AI that works, battle tested - codex - claude code - pi / omp - crush - opencode If... you have 12GB of Nvidia VRAM build a box and joinshow more

Is this the repo?

Yes

I’m trying not to be a doomer but I have 24gb of vram can someone onboard my like I’m a little baby

How? As in what do I download to get started and how do i use it? Already on omarchy, of course.

what actually fits in 12gb that's good enough to code with

Qwen3.8-27b

27b at four bit is already past 12gb before the kv cache, what quant are you running

@0xSero Crippling ~3 bpw quants probably, the fact that pplx evals are missing is sus

@hecodesforme Exl3 is nvfp4 equivalent at 3.5 bits so nah

@frendudebro 3.5 bits on 27b is about 11.8gb of weights, which leaves roughly a gig for context on a 12gb card

I got a 5080 w/ 16GB wish I'd have sprung for the 5090 when I had the chance but even so seems like people are getting good results with on 12GB so it's worth a shot. Really starting to get into agentic coding. Grok Build is so much fun.

Something like this? Hahaha it's not even half!

what about amd cards?

Bar on the bottom - cue the outrage.

Impressive work! The future feels like it is only a few weeks away at this rate

RDNA3 support?

12GB VRAM is tight, check model quantization

12gb makes this a lot more approachable than most local setups i've seen

Which model do you pair with those coding agents on a 12GB card? Curious how it holds up on longer tasks.

24GB VRAM AMD here. Is there hope for me?

#KeepOmarchying

What is the blocker to support MBP’s unified memories?

Man this monochrome color scheme is beautiful

q2 model with super tight context length? No looping issues?

I have an AMD gpu am I invited to the party?

bro just release omarchy for m4

Unless it has some sort of blazing fast, on-metal computer use I don’t know about, this is all hype & slop on top of more slop

What is the best model for gtx 1070 8GB in you opinion? 32gb system ram, Ryzen 9800x3d. Mostly interested in front end development, image editing.

🫶

fvck I have only 11 :(

I am once again asking you to benchmark and improve the B580 experience

nice. I've got a 12GB box sitting idle, what model you running under it?

fye

we gave the local agent the same tool list once and the permission gate blocked a needed write. exact claude code call stopped at the gate, so we logged the route, spent nothing, and left the change unapplied.

Excuse me how are Codex and Claude Code "free AI"

can we get some love for mini-computers, like GMKTec Evo-X1/X2 amd ryzen ai 9 hx*

Finally. Wrangling local LLMs feels like tuning a vintage engine until now. That clean UI is the real magic, letting you focus on what Codex builds instead of how it runs. Smart move keeping your workflow tight while others get lost in config files.

gonna have to add a share compute power function so i can donate my powers to you for training.

will 128gb halo work with gwen3.8 27b?

Could you recommend best settings for Qwen3.8-27b and RTX 5070ti (16GB) please?