local intelligence

can it run

Can the Radeon 680M run qwen2.5vl 32B?

NO

23.0 GB vs 13.5 GB usable — does not fit

~13.5GB usable · RDNA2 (iGPU) · model file 22GB · 32B

~1.2–3.7 tok/s

computed roofline: 102.4 GB/s × 0.25–0.80 efficiency window / 22.0GB (Q4_K_M) — bands, never points.

Run qwen2.5vl 32B on instead

Mac mini M5 Pro (64GB) — cheapest catalog machine that runs qwen2.5vl 32B fully ($1669)

every model the Radeon 680M can run — full list

explore the full catalog84 machines indexed · live prices · what each one can run
2026-09-17 · ← all models on the Radeon 680M · fit = weights + 4k KV vs usable memory