Google

Gemma 4 12B

Google·12B·Gemma·256K ctx·2026-08

Check on my hardwareAdd to compare →Ollama →Hugging Face →

Model size→Hardware requirement→Actual fit

Gemma 4 12B needs ~8.8–15.7 GB of VRAM depending on quantization. The bar is what it needs; the marker is the card's VRAM. Speed figures below are planning estimates.

8 GB

RTX 4060 · 3060 Ti

8.8
8
Q4_K_M · Est. ~3 tok/sSlow · CPU offload

12 GB

RTX 4070 · 3060 12G

10.2
12
Q5_K_M · Est. ~32 tok/sComfortable

16 GB

RTX 4080 · M-series 16G

10.2
16
Q5_K_M · Est. ~50 tok/sComfortable

24 GB

RTX 4090 · 3090

15.7
24
Q8_0 · Est. ~49 tok/sComfortable

48 GB

A6000 · 2×24G

15.7
48
Q8_0 · Est. ~49 tok/sComfortable

02  — Quantizations

QuantFile sizeEst. VRAM
Q4_K_M7.6 GB8.8 GB
Q5_K_M9 GB10.2 GB
Q8_014.5 GB15.7 GB

03  — Runtimes & use

llama.cpp · Ollama · LM Studio

General

Memory and speed figures are planning estimates, not measured benchmarks. Memory includes a fixed allowance for KV cache and runtime overhead. Long-context prompts push VRAM use above the numbers shown here.