local intelligence

can it run

Can the RTX 3090 (24GB, used) run gemma3 12B?

YES

Comfortably. 24GB of memory against a 11GB working set leaves headroom for context, the OS, and anything else you're running.

~53–88 tok/s

weights 8.1GB + KV cache 1.5GB at 4k context vs ~24GB usable → full. 9.6 GB needed of 24 GB usable — headroom for context

computed roofline: 936 GB/s × 0.46–0.76 efficiency window / 8.1GB (Q4_K_M) — bands, never points.

Community check: ~49 tok/s · 12-15B Q4 · 34 runs · inside our computed band — localmaxxing.com

explore the full catalog84 machines indexed · live prices · what each one can run

Specs

model file8.1GB (GGUF Q4_K_M)
minimum memory11GB working set
this machine24GB GDDR6X, 936 GB/s
params12B
licensegemma
price$1100

Run it

model files: hugging face ↗

easy run: ollama pull gemma3:12b

runs with: ollama · llama.cpp · koboldcpp · lm studio

Alternatives

ASUS ExpertCenter Pro ET900N G3 (GB300) — this model on the biggest machine

rtx-5090 — the next step up from this machine's GPU

Buy

NVIDIA (used market) ↗ · Amazon ↗

2026-09-17 · ← full hardware catalog · prices verified at source, may drift