Rent a GPU

Gemma 4 12B hardware requirements

11.96B parameters. Needs about 8.3 GB at Q4_K_M with 8K context.

Parameters12BDense
Q4_K_M · 8K8.3 GBthe usual download
FP16 · 8K24.9 GBfull precision
Runs well on45/51GPUs and Macs we track

Memory needed at each quantization

Weights, KV cache and runtime overhead, at 8K and 32K tokens of context.

QuantQuality8K context32K context
FP16Full quality24.9 GB25.1 GB
Q8_0Near-lossless13.7 GB13.9 GB
Q6_KExcellent10.8 GB11 GB
Q5_K_MVery good9.5 GB9.7 GB
Q4_K_MGood, most popular8.3 GB8.5 GB
Q3_K_MNoticeably degraded6.8 GB7 GB
Q2_KHeavily degraded5.5 GB5.7 GB

Run it with Ollama

Downloads Q4_K_M by default.

ollama run gemma4:12b

Model card

google/gemma-4-12B-it on Hugging Face: licence, prompt format and GGUF builds.

Run it in the cloud

Run Gemma 4 12B on a rented GPU today.

It needs about 8.3 GB at Q4_K_M: a rented RTX 3090 / RTX 4090 runs it for about $0.18 to $0.89 an hour.

RunPodRent a GPU on RunPodVast.aiRent a GPU on Vast.ai

Every GPU and Mac checked

Sorted by memory. Bars show how much of each card Gemma 4 12B fills at the best quant that fits.

45 Runs well

NameMemory usedBest quantSpeed

6 Fits with heavy quantization

NameMemory usedBest quantSpeed

Other Gemma models