can it run
Comfortably. 256GB of memory against a 5GB working set leaves headroom for context, the OS, and anything else you're running.
weights 3.3GB + KV cache 0.53GB at 4k context vs ~32GB usable → full. 3.8 GB needed of 32 GB usable — headroom for context
computed roofline: 1792 GB/s × 0.55–0.80 efficiency window / 3.3GB (Q4_K_M) — bands, never points. Small models (under ~4.5B) run BELOW these bands — kernel-launch overhead keeps them from saturating bandwidth (measured against community medians, 09-06).
| model file | 3.3GB (GGUF Q4_K_M) |
| minimum memory | 5GB working set |
| this machine | 256GB VRAM (8× 32GB) |
| params | 4B |
| license | gemma |
| price | $100000 |
easy run: ollama pull gemma3:4b
runs with: ollama · llama.cpp · koboldcpp · lm studio
no alternatives — every machine in the catalog can run this one.
→ rtx-pro-6000 — the next step up from this machine's GPU