GGUF builds of google/gemma-4-31B-it-assistant — the official Gemma 4 Multi-Token Prediction (MTP) drafter for google/gemma-4-31B-it. Use it as a speculative-decoding draft model alongside the matching Gemma 4 target to get a meaningful decoding speedup at zero quality loss.

Approximate size: 0.5B (assistant) / 33B target.

[!IMPORTANT] These GGUFs use the custom gemma4_assistant architecture and will not load in stock llama.cpp. They require the atomic-llama-cpp-turboquant fork, which adds:

the gemma4_assistant MTP drafter arch (incl. the centroid LM head for E2B/E4B),

TurboQuant KV-cache quantization (-ctk turbo3 -ctv turbo3),

the --mtp-head / --spec-type mtp runtime flags.

Loading these files in upstream ggml-org/llama.cpp will fail with an unknown architecture error.

Files

File	Quant	Size	Notes
`gemma-4-31B-it-assistant.F16.gguf`	F16	910.6 MB	reference (smallest quality loss vs source)
`gemma-4-31B-it-assistant.Q8_0.gguf`	Q8_0	490.8 MB	near-lossless 8-bit
`gemma-4-31B-it-assistant.Q5_K_M.gguf`	Q5_K_M	359.1 MB	balanced k-quant
`gemma-4-31B-it-assistant.Q4_K_M.gguf`	Q4_K_M	337.1 MB	recommended default for speculative-decoding draft
`gemma-4-31B-it-assistant.Q4_K_S.gguf`	Q4_K_S	333.0 MB	smallest k-quant

Quick start

Build the fork:

bash

git clone https://github.com/AtomicBot-ai/atomic-llama-cpp-turboquant
cd atomic-llama-cpp-turboquant
# Pick one of the platform-specific configurations:
cmake -B build -DGGML_METAL=ON          # Apple Silicon
# cmake -B build -DGGML_CUDA=ON         # NVIDIA
# cmake -B build                        # CPU-only
cmake --build build --target llama-server llama-cli llama-quantize -j

Download the assistant drafter (this repo) and the matching Gemma 4 target:

bash

hf download AtomicChat/gemma-4-31B-it-assistant-GGUF \
    --include "*Q4_K_M.gguf" --local-dir ./models
# Any GGUF build of the matching target model works; e.g. unsloth's:
hf download unsloth/gemma-4-31B-it-GGUF \
    --include "*Q4_K_M*.gguf" --local-dir ./models

Run llama-server with MTP speculative decoding + TurboQuant KV cache:

bash

./build/bin/llama-server \
    -m         ./models/gemma-4-31B-it-Q4_K_M.gguf \
    --mtp-head ./models/gemma-4-31B-it-assistant.Q4_K_M.gguf \
    --spec-type mtp \
    --draft-block-size 3 --draft-max 8 --draft-min 0 \
    -ngl 99 -ngld 99 \
    -ctk turbo3 -ctv turbo3 -ctkd turbo3 -ctvd turbo3 \
    -fa on -c 16384 --host 127.0.0.1 --port 8080

A ready-made launcher lives at scripts/run-gemma4-31b-mtp-server.sh in the fork (MTP_PRESET=throughput|lift|balanced|quality).

How MTP works here

Gemma 4 ships with a small "assistant" head that predicts several future tokens from the target model's last hidden state. In atomic-llama-cpp-turboquant it is loaded as a separate GGUF via --mtp-head and drives a custom speculative decoder (block_size 3, draft_max 16 for quality preset). The verifier runs the target model in parallel, guaranteeing the same output distribution as plain greedy/sampled decoding.

TurboQuant KV cache

turbo3 is the KV-cache quantization scheme used in this fork; it significantly reduces KV memory and bandwidth at long contexts with no measurable quality regression on Gemma 4. Apply it to both target and drafter via -ctk turbo3 -ctv turbo3 -ctkd turbo3 -ctvd turbo3.

License & attribution

Released under the Gemma Terms of Use.

Original model card: google/gemma-4-31B-it-assistant
Target model: google/gemma-4-31B-it
License text: https://ai.google.dev/gemma/docs/gemma_4_license

Acknowledgements

Google DeepMind — Gemma 4 family and the MTP drafters.
ggml-org/llama.cpp — upstream inference engine.
TurboQuant primitives — KV-cache quantization scheme integrated in the fork.

— Atomic Chat

Phil2Sat/gemma-4-31B-it-assistant-GGUF

Gemma 4 31B Assistant — GGUF (Atomic Chat)

Files

Quick start

How MTP works here

TurboQuant KV cache

License & attribution

Acknowledgements

Phil2Sat/
gemma-4-31B-it-assistant-GGUF