local intelligence

can it run

Can the NVIDIA GeForce RTX 4080 SUPER run qwen3 32B?

NO

21.0 GB vs 16 GB usable — does not fit

16GB · Ada · model file 20GB · 32B

~18–29 tok/s

computed roofline: 736 GB/s × 0.50–0.80 efficiency window / 20.0GB (Q4_K_M) — bands, never points.

Run qwen3 32B on instead

Mac mini M4 (32GB) — cheapest catalog machine that runs qwen3 32B fully ($999)

rtx-4090 — the next step up from NVIDIA GeForce RTX 4080 SUPER

every model the NVIDIA GeForce RTX 4080 SUPER can run — full list

explore the full catalog84 machines indexed · live prices · what each one can run
2026-09-17 · ← all models on the NVIDIA GeForce RTX 4080 SUPER · fit = weights + 4k KV vs usable memory