can it run
Comfortably. 128GB of memory against a 28GB working set leaves headroom for context, the OS, and anything else you're running.
weights 22GB + KV cache 1GB at 4k context vs ~126GB usable → full. 23.0 GB needed of 126 GB usable — headroom for context
computed roofline: 273 GB/s × 0.25–0.65 efficiency window / 22.0GB (Q4_K_M) — bands, never points.
| model file | 22GB (GGUF Q4_K_M) |
| minimum memory | 28GB working set |
| this machine | 128GB unified LPDDR5x |
| params | 32B |
| license | apache-2.0 |
| price | $3900 |
easy run: ollama pull qwen2.5vl:32b
runs with: ollama · llama.cpp · koboldcpp · lm studio
→ ASUS ExpertCenter Pro ET900N G3 (GB300) — this model on the biggest machine
→ rtx-5090 — the next step up from this machine's GPU