GE

Gemma-4-12B-QAT-Heretic-StyleTune

GemmaGGUF by motionsilseUpdated 2026-09-05
Download on Hugging Face
Parameters12.9B
Context256K
Layers48
Architecturegemma4
Smallest quant8.0 GB
Quick answer

Gemma-4-12B-QAT-Heretic-StyleTune needs about 21.4 GB of memory at Q4_0 (a 8.0 GB file) with an 8K context window, so it fits on a 24 GB GPU. Longer context needs more: at 32K tokens the KV cache alone adds about 51.5 GB.

Gemma‑4‑12B‑QAT‑Heretic‑StyleTune – a 12B Gemma model with a style‑tuned output head

Gemma‑4‑12B‑QAT‑Heretic‑StyleTune is a 12.9B‑parameter Gemma model with a single‑tensor style tune. It retains all reasoning and tool‑use capabilities while providing a distinct narrative voice. The model is available as a GGUF file quantized to Q4_0.

Good at

  • Reasoning
  • Tool use
  • Narrative style

Limits

  • No vision support
  • Requires ~21.4 GB VRAM at 8 k context
  • Single‑tensor style tune only

Best for

  • Text generation with a unique voice
  • Fine‑tuning only the lm_head
  • Running on high‑VRAM GPUs

Who it’s for. Developers and researchers who need a Gemma model with a distinct narrative style but want to keep training costs low. The model is suitable for applications that rely on reasoning and tool use without requiring multimodal input.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q4_08.0 GB 21.4 GB
Too big Get file →

The only available quant is Q4_0, which results in a file size of 8.05 GB and an estimated VRAM usage of 21.4 GB at 8 k context. For lower VRAM budgets, this model is not recommended. If you have at least 22 GB of VRAM, Q4_0 is the appropriate choice.

What GPU do you need for Gemma-4-12B-QAT-Heretic-StyleTune?

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: no quant fits at an 8K context.
  • 16 GB VRAM: no quant fits at an 8K context.
  • 24 GB VRAM: the best quant that fits is Q4_0 (8.0 GB file, about 21.4 GB in use).

Which Gemma-4-12B-QAT-Heretic-StyleTune quant should you choose?

Q4_0 (8.0 GB) is the middle option in this repository. 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

Architecturegemma4
Parameters12.9B
Layers48
Trained context262,144 tokens
KV cache (fp16)1536 KB per token

Source

Quantized bymotionsilse
FamilyGemma
Licenseapache-2.0
Downloads (HF)242
Last updated2026-09-05

Frequently asked questions about Gemma-4-12B-QAT-Heretic-StyleTune

How much VRAM does Gemma-4-12B-QAT-Heretic-StyleTune need?

About 21.4 GB at Q4_0 with an 8K context window. That is the 8.0 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 Gemma-4-12B-QAT-Heretic-StyleTune on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 21.4 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Gemma-4-12B-QAT-Heretic-StyleTune GGUF?

It depends on the quant: Q4_0 is 8.0 GB.

What context length does Gemma-4-12B-QAT-Heretic-StyleTune support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 12.9 GB at 8K and 51.5 GB at 32K.

Where can I download Gemma-4-12B-QAT-Heretic-StyleTune GGUF?

The files are on Hugging Face in the motionsilse/Gemma-4-12B-QAT-Heretic-StyleTune-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 Gemma-4-12B-QAT-Heretic-StyleTune use?

Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run this model on 8 GB VRAM?

No. The estimated VRAM usage at 8 k context is 21.4 GB, so 8 GB is insufficient.

Does this model support vision or multimodal input?

No. The model only supports text generation and does not include a multimodal projector.

What is the maximum context length for this model?

The trained context length is 262144 tokens.

Other GGUFs of Gemma-4-12B-StyleTune

Similar models

Browse all →

More Gemma models

All Gemma →

Other medium models

All medium models →