local intelligence

can it run

Can the NVIDIA GeForce RTX 5070 run gemma3 4B?

FULL

3.8 GB needed of 12 GB usable — headroom for context

12GB · Blackwell · model file 3.3GB · 4B

~110–160 tok/s

computed roofline: 672 GB/s × 0.56–0.80 efficiency window / 3.3GB (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).

Community check: ~180 tok/s · 3-4B Q4 · 3 runs · inside our computed band — localmaxxing.com

Run gemma3 4B on instead

Raspberry Pi 5 + AI Kit (Hailo-8L) — cheapest catalog machine that runs gemma3 4B fully ($110)

rtx-5070-ti — the next step up from NVIDIA GeForce RTX 5070

every model the NVIDIA GeForce RTX 5070 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 5070 · fit = weights + 4k KV vs usable memory