local intelligence

can it run

Can the NVIDIA GeForce RTX 3080 12GB run LLAMA4 Scout 109B A17B?

NO

67.8 GB vs 12 GB usable — does not fit

12GB · Ampere · model file 67GB · Scout 109B A17B

~6.3–10 tok/s

computed roofline: 912.4 GB/s × 0.46–0.76 efficiency window / 67.0GB (Q4_K_M) — bands, never points.

Run LLAMA4 Scout 109B A17B on instead

NIMO AI Mini PC (128GB) — cheapest catalog machine that runs LLAMA4 Scout 109B A17B fully ($1999)

rtx-3090 — the next step up from NVIDIA GeForce RTX 3080 12GB

every model the NVIDIA GeForce RTX 3080 12GB 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 3080 12GB · fit = weights + 4k KV vs usable memory