Gemma 3 on NVIDIA GeForce RTX 4070 · 12GB VRAM

✅ Runs (tight) — Gemma 3 12B @ Q4_K_M
Gemma 3 12B at Q4_K_M needs ~10.4 GB (weights 6.9 GB + KV 3.0 GB + overhead 512 MB @ 8K ctx) of your 10.8 GB usable VRAM — fits, but little headroom — close other apps or trim context. Expect ~20–35 tok/s (fast).
context @ Q4_K_M: comfortable to 2K · runs to 32K · won't fit past that on this rig
quantneedsspeed
~ Q8_015.7 GB~4–7 tok/s (slow)
~ Q5_K_M11.6 GB~6–9 tok/s (slow)
✓ Q4_K_M10.4 GB~20–35 tok/s (fast)
✓ Q3_K_M9.0 GB~25–40 tok/s (fast)
$runlocal install gemma3:12b
⚠ NVIDIA support is best-effort in v0.1 — verify before relying on it.
runlocal verdict card
download card share on x share on reddit drop it in your model-drop post
Gemma 3: Runs (tight)
for your own README — links back here

Same model, other machines: Apple M1 · Apple M1 Pro · Apple M2 · Apple M3 Pro · Apple M2 Pro · Apple M4 Pro

Also runs on a NVIDIA GeForce RTX 4070: Qwen3 · DeepSeek-R1 (distill) · Qwen2.5 · Llama 3.1 · Gemma 4

Best models for a NVIDIA GeForce RTX 4070 · 12GB VRAM →

Architecturally similar

These share Gemma 3's KV-cache geometry — they size memory the same way, so here's how they fit on a NVIDIA GeForce RTX 4070 · 12GB VRAM too:

modelsharesverdict here
Gemma 2 16kv/128hd4kv/256hd8kv/256hd ✅ Runs (tight)
Falcon3 7B 4kv/256hd ✅ Runs (tight)

Check any combo yourself: open the checker →