No changes to datasets or capabilities — fully functional, 100% of what the original authors intended, just without the refusals. Built from the official QAT weights, so the 4-bit quant stays close to full-precision quality.

Balanced

The Balanced variant (recommended — 99%+ of users will be happy here) uses optimized full uncensoring tuned especially for agentic coding, reasoning, creative writing and reliability-critical tasks. It reasons before answering and stays dependable and on-instruction. An Aggressive variant, for cases where Balanced still deflects too much, after current testing is not required.

~53% faster with MTP

Ships with an MTP (multi-token-prediction) draft head for speculative decoding — roughly 53% faster generation with identical output (the model verifies every drafted token, so quality is unchanged — pure speed). This release is tuned to pair well with the included MTP head.

llama.cpp:

bash

llama-server \
  -m Gemma4-31B-QAT-Uncensored-HauhauCS-Balanced-Q4_K_M.gguf \
  -md mtp-gemma-4-31B-it.gguf --spec-type draft-mtp \
  -ngl 99 -fa on

Note: the MTP speedup was currently tested by me through llama.cpp (llama-server / llama-cli).

Downloads

File	Type	Size
`Gemma4-31B-QAT-Uncensored-HauhauCS-Balanced-Q4_K_M.gguf`	Q4_K_M (text)	18.7 GB
`mmproj-Gemma4-31B-QAT-Uncensored-HauhauCS-Balanced-BF16.gguf`	mmproj (vision)	1.2 GB
`mtp-gemma-4-31B-it.gguf`	MTP speculative drafter	280 MB

Why only Q4_K_M? Gemma 4 is quantization-aware-trained for ~4-bit, so Q4_K_M is the sweet spot — higher-precision quants add size with no real quality gain. Carefully quantized for best quality at 4-bit.

Vision

Load the mmproj alongside the model for image input:

bash

llama-server -m Gemma4-31B-QAT-Uncensored-HauhauCS-Balanced-Q4_K_M.gguf \
  --mmproj mmproj-Gemma4-31B-QAT-Uncensored-HauhauCS-Balanced-BF16.gguf -ngl 99 -fa on

Recommended sampling

These are dialed in specifically for this HauhauCS build — use them for the intended behaviour and quality:

temperature 0.6
top_k 64
top_p 0.9
min_p 0.05
repeat_penalty 1.1

This release is tuned end-to-end as its own thing; the settings above are part of that and aren't the stock Gemma defaults.

Specs

31B dense · 256K (262144) context
Vision (image input) via mmproj
Based on Gemma 4 31B by Google DeepMind

Compatibility

Works with llama.cpp, LM Studio, Jan, koboldcpp, and other GGUF runtimes.
Multi-GPU + LM Studio: I've personally noticed Gemma 4 can crash under LM Studio's tensor-split mode — use a single GPU (layer-split or priority order) for this model.

Acknowledgements

Google DeepMind — Gemma 4.
The included mtp-gemma-4-31B-it.gguf speculative draft head comes from Unsloth's Gemma 4 release — many thanks to the Unsloth team for it.

* Tested with both automated and manual refusal benchmarks — none have been found in standard use. A small number of edge-case prompts deflect on the first ask but comply on a re-ask or strategic framing. If you hit one that's actually obstructive to your use case, join the Discord and flag it so I can work on it in a future revision.

HauhauCS/Gemma4-31B-QAT-Uncensored-HauhauCS-Balanced-MTP

Gemma4-31B-QAT-Uncensored-HauhauCS-Balanced-MTP

About