local intelligence

can it run

Can the Radeon PRO W7900 48GB run qwen3.8 27B?

FULL

18.5 GB needed of 48 GB usable — headroom for context

48GB · RDNA3 · model file 16.5GB · 27B

~26–39 tok/s

computed roofline: 864 GB/s × 0.50–0.75 efficiency window / 16.5GB (Q4_K_M) — bands, never points.

Run qwen3.8 27B on instead

Tesla V100 32GB (used) — cheapest catalog machine that runs qwen3.8 27B fully ($640)

rtx-pro-6000 — the next step up from Radeon PRO W7900 48GB

every model the Radeon PRO W7900 48GB can run — full list

explore the full catalog90 machines indexed · live prices · what each one can run
2026-09-24 · ← all models on the Radeon PRO W7900 48GB · fit = weights + 4k KV vs usable memory