# gemma-3-12b-it GPU requirements

gemma-3-12b-it has 12B parameters. To serve it, plan for about 29 GB of GPU memory in 16-bit, 15 GB in 8-bit or 7.3 GB in 4-bit, including about 20% for the KV cache and activations.

[Deploy it on Arrakis](/console/#/models?model=google/gemma-3-12b-it).

## What it takes to serve it

| Precision | Memory and the smallest setup |
|---|---|
| 16-bit | 29 GB: one [RTX 5090](/gpus/rtx-5090.html.md), from $0.58 per GPU-hour |
| 8-bit | 15 GB: one [RTX 4090](/gpus/rtx-4090.html.md) or [RTX 3090](/gpus/rtx-3090.html.md), RTX 3090 from $0.19 per GPU-hour |
| 4-bit | 7.3 GB: one [RTX 4090](/gpus/rtx-4090.html.md) or [RTX 3090](/gpus/rtx-3090.html.md), RTX 3090 from $0.19 per GPU-hour |

## What it takes to fine-tune it

| Method | Memory and the smallest setup |
|---|---|
| QLoRA | 8.5 GB: one [RTX 4090](/gpus/rtx-4090.html.md) or [RTX 3090](/gpus/rtx-3090.html.md), RTX 3090 from $0.19 per GPU-hour |
| LoRA | 29 GB: one [RTX 5090](/gpus/rtx-5090.html.md), from $0.58 per GPU-hour |
| Full fine-tuning | 195 GB: one [B300](/gpus/b300.html.md), from $4.60 per GPU-hour |

## About the model

| | |
|---|---|
| Parameters | `12B` |
| Published in | `BF16` |
| License | `gemma` |
| Released | `2025-03-01` |
| Weights | [google/gemma-3-12b-it](https://huggingface.co/google/gemma-3-12b-it) |

## Serve it yourself

```
vllm serve google/gemma-3-12b-it --tensor-parallel-size 1
```

## Questions

**How much GPU memory does gemma-3-12b-it need?** To serve it, about 29 GB in 16-bit, 15 GB in 8-bit, 7 GB in 4-bit, including about 20% for the KV cache and activations. To fine-tune it, about 9 GB with QLoRA, 29 GB with LoRA and 195 GB for full fine-tuning with Adam, before activations.

**Can I run gemma-3-12b-it on one GPU?** Yes. In 16-bit it needs about 29 GB, which fits on one RTX 5090.

**What does it cost to run gemma-3-12b-it?** In 16-bit it needs one RTX 5090; in 4-bit, one RTX 4090 or RTX 3090. Each setup on this page shows the lowest price per GPU-hour on Arrakis right now, and you pay by the hour from a prepaid balance.

**How do I deploy gemma-3-12b-it on Arrakis?** Open Models in the Arrakis console, paste google/gemma-3-12b-it, choose vLLM, SGLang or llama.cpp, and rent a GPU it fits on. You get an OpenAI-compatible endpoint with its own API key, and root SSH to the machine.


More models: [/models](/models.html.md).