local intelligence

can it run

Can the Radeon 680M run LLAMA4 Scout 109B A17B?

NO

67.8 GB vs 13.5 GB usable — does not fit

~13.5GB usable · RDNA2 (iGPU) · model file 67GB · Scout 109B A17B

~0.38–1.2 tok/s

computed roofline: 102.4 GB/s × 0.25–0.80 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)

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