can it run
Comfortably. 32GB of memory against a 7GB working set leaves headroom for context, the OS, and anything else you're running.
weights 5.6GB + KV cache 0.22GB at 4k context vs ~30GB usable → full. 5.8 GB needed of 30 GB usable — headroom for context
computed roofline: 120 GB/s × 0.80–1.00 efficiency window / 5.6GB (Q4_K_M) — bands, never points.
What users report — 3 cited rows:
→ 23 tok/s · Llama 3.1 8B Q4_K_M Ollama/Metal · source
→ 18.15 tok/s · Llama 3.1 8B Q5_K_M Ollama/Metal (primary run) · source
→ ~20 tok/s · 8B Q4_K_M aggregated estimate · source
note: optimized engines (MLX / MTP / speculative decode) measured ×1.3–1.9 over the base stack on 8B-class — a note, not a second band. ollama.com · reddit.com · genie.devoxx.com
→explore the full catalog84 machines indexed · live prices · what each one can run| model file | 5.6GB (GGUF Q4_K_M) |
| minimum memory | 7GB working set |
| this machine | 32GB unified |
| params | 7B |
| license | apache-2.0 |
| price | $999 |
easy run: ollama pull qwen2.5vl:7b
runs with: ollama · llama.cpp · koboldcpp · lm studio
no alternatives — every machine in the catalog can run this one.