GE

Google_Gemma 3 12b It

GemmaGGUF by marceloneUpdated 2025-07-31
Download on Hugging Face
Parameters11.8B
Context128K
Layers48
Architecturegemma3
Smallest quant6.9 GB
Quick answer

Google_Gemma 3 12b It needs about 16.2 GB of memory at Q8_0 (a 12.5 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 6 quants, from 6.9 GB (IQ4_NL) to 47.1 GB (F32). Longer context needs more: at 32K tokens the KV cache alone adds about 12.9 GB.

Google_Gemma 3 12b It is a 11.8B-parameter Gemma model in GGUF format, quantized by marcelone. Its trained context length is 128K tokens. The repository has 6 quant files, from 6.9 GB (IQ4_NL) to 47.1 GB (F32).

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ4_NL6.9 GB 10.6 GB
Too big Get file →
Q6_K9.7 GB 13.4 GB
Too big Get file →
Q8_012.5 GB 16.2 GB
Too big Get file →
BF1623.5 GB 27.3 GB
Too big Get file →
F1623.5 GB 27.3 GB
Too big Get file →
F3247.1 GB 50.8 GB
Too big Get file →

What GPU do you need for Google_Gemma 3 12b It?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: no quant fits at an 8K context.
  • 6 GB VRAM: no quant fits at an 8K context.
  • 8 GB VRAM: no quant fits at an 8K context.
  • 12 GB VRAM: the best quant that fits is IQ4_NL (6.9 GB file, about 10.6 GB in use).
  • 16 GB VRAM: the best quant that fits is Q6_K (9.7 GB file, about 13.4 GB in use).
  • 24 GB VRAM: the best quant that fits is Q8_0 (12.5 GB file, about 16.2 GB in use).
  • 32 GB VRAM: the best quant that fits is BF16 (23.5 GB file, about 27.3 GB in use).
  • 48 GB VRAM: the best quant that fits is BF16 (23.5 GB file, about 27.3 GB in use).

Which Google_Gemma 3 12b It quant should you choose?

Q8_0 (12.5 GB) is the middle option in this repository. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.

If memory is tight, the smallest file is IQ4_NL at 6.9 GB, with the quality trade-offs that come with fewer bits. The usual sweet spot between file size and quality, and the quant most people start with.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturegemma3
Parameters11.8B
Layers48
Trained context131,072 tokens
KV cache (fp16)384 KB per token

Source

Quantized bymarcelone
FamilyGemma
Licensegemma
Downloads (HF)287
Last updated2025-07-31

Frequently asked questions about Google_Gemma 3 12b It

How much VRAM does Google_Gemma 3 12b It need?

About 16.2 GB at Q8_0 with an 8K context window. That is the 12.5 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run Google_Gemma 3 12b It on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 10.6 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Google_Gemma 3 12b It GGUF?

It depends on the quant: IQ4_NL is 6.9 GB, Q6_K is 9.7 GB, Q8_0 is 12.5 GB, BF16 is 23.5 GB, F16 is 23.5 GB, F32 is 47.1 GB.

What context length does Google_Gemma 3 12b It support?

The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 3.2 GB at 8K and 12.9 GB at 32K.

Where can I download Google_Gemma 3 12b It GGUF?

The files are on Hugging Face in the marcelone/gemma-3-12b-it-gguf repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does Google_Gemma 3 12b It use?

Hugging Face lists the license as gemma. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Other GGUFs of gemma-3-12b-it

Similar models

Browse all →

More from marcelone

All by marcelone →

More Gemma models

All Gemma →

Other medium models

All medium models →