can it run
Comfortably. 80GB of memory against a 25GB working set leaves headroom for context, the OS, and anything else you're running.
weights 20GB + KV cache 1GB at 4k context vs ~80GB usable → full. 21.0 GB needed of 80 GB usable — headroom for context
computed roofline: 2039 GB/s × 0.46–0.76 efficiency window / 20.0GB (Q4_K_M) — bands, never points.
| model file | 20GB (GGUF Q4_K_M) |
| minimum memory | 25GB working set |
| this machine | 80GB HBM2e, 2 TB/s |
| params | distill 32B |
| license | mit |
| price | $23000 |
easy run: ollama pull deepseek-r1:32b
runs with: ollama · llama.cpp · koboldcpp · lm studio
→ ASUS ExpertCenter Pro ET900N G3 (GB300) — this model on the biggest machine
→ rtx-pro-6000 — the next step up from this machine's GPU
NVIDIA (used market) ↗ · Amazon ↗