local intelligence

can it run

Can the NVIDIA L40S 48GB run gemma3 1B?

FULL

1.1 GB needed of 48 GB usable — headroom for context

48GB · Ada Lovelace · model file 1.0GB · 1B

~400–660 tok/s

computed roofline: 864 GB/s × 0.46–0.76 efficiency window / 1.0GB (Q4_K_M) — bands, never points. Small models (under ~4.5B) run BELOW these bands — kernel-launch overhead keeps them from saturating bandwidth (measured against community medians, 09-06).

Run gemma3 1B on instead

Tesla M60 (16GB, used) — cheapest catalog machine that runs gemma3 1B fully ($35)

rtx-pro-6000 — the next step up from NVIDIA L40S 48GB

every model the NVIDIA L40S 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 NVIDIA L40S 48GB · fit = weights + 4k KV vs usable memory