local intelligence

can it run

Can the Apple M4 run qwen2.5vl 7B?

FULL

5.8 GB needed of 14 GB usable — headroom for context

~14GB usable · Apple M4 · model file 5.6GB · 7B

~17–21 tok/s

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

Run qwen2.5vl 7B on instead

Tesla P100 (16GB, used) — cheapest catalog machine that runs qwen2.5vl 7B fully ($135)

every model the Apple M4 can run — full list

explore the full catalog84 machines indexed · live prices · what each one can run
2026-09-18 · ← all models on the Apple M4 · fit = weights + 4k KV vs usable memory