can it run
Comfortably. 128GB 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 ~126GB usable → full. 21.0 GB needed of 126 GB usable — headroom for context
computed roofline: 256 GB/s × 0.60–0.85 efficiency window / 20.0GB (Q4_K_M) — bands, never points.
| model file | 20GB (GGUF Q4_K_M) |
| minimum memory | 25GB working set |
| this machine | 128GB LPDDR5X-8000 |
| params | distill 32B |
| license | mit |
| price | $3199 |
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