can it run
Only with compromises. 192GB is below the 212GB working set — expect a lower quant, offloading, or heavy swap. Possible, not pleasant.
weights 168.9GB + KV cache 2GB at 4k context vs ~190GB usable → tight. 170.9 of 190 GB usable — keep context modest
computed roofline: 256 GB/s × 0.60–0.85 efficiency window / 25.3GB (Q4_K_M) — bands, never points.
MoE: file is 168.9GB (that decides fit) but only ~82.8B of experts activate per token (~25.3GB streamed) — band below is the active-expert estimate, optimistic; community MoE rows run several× the naive bandwidth math
| model file | 168.9GB (GGUF mixed-Q2) |
| minimum memory | 212GB working set |
| this machine | 192GB unified (160GB VRAM partition) |
| params | 552B+196B Engram MoE |
| license | MIT |
| price | $5000 |
easy run: no stock runtime yet
→ ASUS ExpertCenter Pro ET900N G3 (GB300) — cheapest machine that runs it ($60000)
→ ASUS ExpertCenter Pro ET900N G3 (GB300) — this model on the biggest machine