gemma-4-26B-A4B ยท GGUF
GGUF builds of google/gemma-4-26B-A4B for llama.cpp / Ollama / LM Studio. Converted by axteam with llama.cpp (converted from the BF16 original). This is a base (pre-trained) model with no chat template โ use plain text completion.
| File | Quant | Size | Notes |
|---|---|---|---|
gemma-4-26B-A4B-GGUF-BF16.gguf |
BF16 | 50.5 GB | unquantized 16-bit (reference; for re-quantizing or max quality) |
gemma-4-26B-A4B-GGUF-Q4_K_M.gguf |
Q4_K_M | 16.8 GB | 4-bit, the usual size/quality trade-off โ the most common choice |
gemma-4-26B-A4B-GGUF-Q8_0.gguf |
Q8_0 | 26.9 GB | 8-bit, closest to BF16 (larger) |
Q4_K_M is the usual default (size/quality trade-off). Q8_0 stays closest to BF16. BF16 is the unquantized reference. Measured perplexity is in the table below. Architecture: Gemma 4 ยท Mixture-of-Experts (multimodal; the GGUF text model runs the language side).
Quantization method
These GGUF files are post-training quantization (PTQ) made with llama.cpp โ the BF16 weights are quantized block-wise (llama.cpp k-quants: per-block scales with mixed bit-width allocation, a refined round-to-nearest), with no retraining. This is a different toolchain from the AWQ / NVFP4 vLLM builds, and it is not QAT (quantization-aware training, which fine-tunes the weights). Q8_0 stays closest to BF16, Q4_K_M is the usual size/quality trade-off (see the measured perplexity below), and BF16 is the unquantized reference for re-quantizing.
Measured on GPU (llama.cpp)
llama.cpp b12005 (commit e0f142d, CUDA build) on an NVIDIA B200, 2026-11-02, all layers on the GPU: perplexity over 16 windows of 2,048 tokens from the WikiText-2 test set and from a Korean Wikipedia sample (lower is better). Percentages are relative to the BF16 file in this repo.
| File | WikiText-2 PPL | Korean wiki PPL | vs BF16 (EN / KO) |
|---|---|---|---|
gemma-4-26B-A4B-GGUF-BF16.gguf |
5.02 | 3.46 | reference |
gemma-4-26B-A4B-GGUF-Q8_0.gguf |
5.02 | 3.47 | +0.0% / +0.1% |
gemma-4-26B-A4B-GGUF-Q4_K_M.gguf |
5.25 | 3.63 | +4.6% / +4.8% |
16 windows per text is a small sample: differences of a few percent between files are within its sampling noise (re-measuring two repos with 64 windows moved such differences by up to ~19 points).
Requirements
| llama.cpp | build โฅ b8637 โ architecture gemma4 needs โฅ b8637, and the commands on this card use -fa on (โฅ b6325). Run here on 2026-11-02 with b12005 (commit e0f142d, CUDA build) on an NVIDIA B200: every file in this repo loaded with all layers on the GPU, its perplexity was measured (see Measured on GPU above) and a 16-token greedy text-completion check was run (base model, no chat template): every file continued โ๋ํ๋ฏผ๊ตญ์ ์๋๋โ with โ์์ธ์ด๋ค. โฆโ, but continued โThe capital of France isโ with โa city of romance, art, and culture. It is also a city of โฆโ without naming Paris. |
| Memory | the chosen GGUF file (sizes in the table) plus KV cache for the context (-c); --cache-type-k/v q8_0 -fa on halves the KV cache. |
| Input | text only โ this repo has no mmproj (vision projector) file. |
| Ollama / LM Studio | not tested here. |
Run with llama.cpp
# download one file (e.g. Q4_K_M) then:
llama-server -m gemma-4-26B-A4B-GGUF-Q4_K_M.gguf -c 8192
# or raw text completion (no chat template):
llama-completion -m gemma-4-26B-A4B-GGUF-Q4_K_M.gguf -no-cnv -p "๋ํ๋ฏผ๊ตญ์ ์๋๋?" -n 128
Reduce KV-cache memory (optional, run-time)
KV-cache type is a run-time choice in llama.cpp (not baked into the file). Halve it with:
llama-server -m gemma-4-26B-A4B-GGUF-Q4_K_M.gguf --cache-type-k q8_0 --cache-type-v q8_0 -fa on
Ollama
ollama run hf.co/axteam/gemma-4-26B-A4B-GGUF:Q4_K_M
Recommended sampling (per Google):
temperature=1.0,top_p=0.95,top_k=64.
License
Quantized derivative of Google's Gemma 4 โ governed by the Gemma 4 license. Weights and tokenizer belong to Google DeepMind.
ํ๊ตญ์ด ์์ฝ
google/gemma-4-26B-A4B์ GGUF ์์ํ ๋น๋ (llama.cppยทOllamaยทLM Studio์ฉ) ยท ์ ์ axteam
๊ฐ์
| ํญ๋ชฉ | ๋ด์ฉ |
|---|---|
| ์๋ณธ ๋ชจ๋ธ | google/gemma-4-26B-A4B |
| ํ์ | GGUF (llama.cpp) |
| ๋ณํ ๋๊ตฌ | llama.cpp convert_hf_to_gguf.py โ llama-quantize |
| ๋ชจ๋ธ ๊ตฌ์กฐ | Gemma 4 ยท MoE (์๋ณธ์ ๋ฉํฐ๋ชจ๋ฌ, GGUF๋ ์ธ์ด ๋ชจ๋ธ ๋ถ๋ถ๋ง ์คํ) |
| ๋ชจ๋ธ ์ ํ | ์ฌ์ ํ์ต base ๋ชจ๋ธ โ chat template ์์, ํ ์คํธ ์ด์ด์ฐ๊ธฐ ์ ์ฉ |
| ํ์ผ ์ | 3๊ฐ |
์์ํ ์์ธ
- ์์ํ ์ ์ฐจ: PTQ(ํ์ต ํ ์์ํ, ์ฌํ์ต ์์), llama.cpp ๋ธ๋ก ๋จ์ ์์ํ(k-quant
Q*_K,Q8_0) - chat template: ์์ โ ์ฌ์ ํ์ต base ๋ชจ๋ธ, ํ ์คํธ ์ด์ด์ฐ๊ธฐ ์ ์ฉ
- ํ์ง(์ค์ธก, 2026-11-02): llama.cpp PPL(WikiText-2ยทํ๊ตญ์ด ์ํค ํ๋ณธ, 2,048ํ ํฐร16, B200 GPU, ๋ฎ์์๋ก ์ข์, ๋น๊ต ๊ธฐ์ค์ ๋ฆฌํฌ์ BF16 ํ์ผ) โ
gemma-4-26B-A4B-GGUF-BF16.gguf์์ด 5.02ยทํ๊ตญ์ด 3.46 ยทgemma-4-26B-A4B-GGUF-Q8_0.gguf์์ด 5.02ยทํ๊ตญ์ด 3.47(BF16 ๋๋น +0.0%ยท+0.1%) ยทgemma-4-26B-A4B-GGUF-Q4_K_M.gguf์์ด 5.25ยทํ๊ตญ์ด 3.63(BF16 ๋๋น +4.6%ยท+4.8%) โ ํ ์คํธ๋น 16์ฒญํฌ๋ผ ํ์ผ ์ฌ์ด ๋ช % ์ฐจ์ด๋ ํ๋ณธ ํ๋ค๋ฆผ ๋ฒ์
์ ๊ณต ํ์ผ
| ํ์ผ | ํฌ๊ธฐ | ์ค๋ช |
|---|---|---|
gemma-4-26B-A4B-GGUF-BF16.gguf |
50.5 GB | ์์ํํ์ง ์์ 16๋นํธ โ ๊ธฐ์ค ํ์ผ (์ฌ์์ํ์ฉยท์ต๊ณ ํ์ง) |
gemma-4-26B-A4B-GGUF-Q4_K_M.gguf |
16.8 GB | 4๋นํธ โ ํฌ๊ธฐยทํ์ง ๊ท ํ, ๊ฐ์ฅ ๋ง์ด ์ฐ๋ ์ ํ |
gemma-4-26B-A4B-GGUF-Q8_0.gguf |
26.9 GB | 8๋นํธ โ BF16์ ๊ฐ์ฅ ๊ฐ๊น์, ์ฉ๋ ํผ |
์๊ตฌ์ฌํญ
- llama.cpp: b8637 ์ด์ (
gemma4๊ตฌ์กฐ ์ง์ b8637๋ถํฐ, ์์ ๋ช ๋ น์-fa on์ต์ b6325๋ถํฐ) - ์คํ ํ์ธ: 2026-11-02 b12005(์ปค๋ฐ e0f142d) CUDA ๋น๋๋ก NVIDIA B200์์ ์ด ๋ฆฌํฌ์ ๋ชจ๋ ํ์ผ์ ์ ์ธต GPU์ ์ฌ๋ ค ์คํ โ PPL ์ธก์ (์ ์์ด ๋ณธ๋ฌธ ํ), 16ํ ํฐ ์ด์ด์ฐ๊ธฐ ์ ๊ฒ(base ๋ชจ๋ธ, chat template ์์) โ ๋ชจ๋ ํ์ผ์ด โ๋ํ๋ฏผ๊ตญ์ ์๋๋โ์ โ์์ธ์ด๋ค. โฆโ๋ก ์ด์๊ณ , โThe capital of France isโ๋ Paris๋ฅผ ๋งํ์ง ์๊ณ โa city of romance, art, and culture. It is also a city of โฆโ๋ก ์ด์
- ๋ฉ๋ชจ๋ฆฌ: ์ ํํ GGUF ํ์ผ ํฌ๊ธฐ + ์ปจํ
์คํธ(
-c)๋งํผ์ KV ์บ์ (KV ์บ์ ์ค์ด๋ ๋ฐฉ๋ฒ์ ์๋ ์ฌ์ฉ๋ฒ ์ฐธ๊ณ ) - ์
๋ ฅ: ํ
์คํธ ์ ์ฉ โ ๋น์ ์ธ์ฝ๋ยทํฌ์์ธต ํ์ผ(
mmproj)์ด ์์ด ์ด๋ฏธ์ง ์ ๋ ฅ ๋ถ๊ฐ - OllamaยทLM Studio: ์ ์ ์ธก์์ ์คํ ๋ฏธํ์ธ
์ฌ์ฉ๋ฒ
- ํ์ผ ๋ฐ๊ธฐ:
hf download axteam/gemma-4-26B-A4B-GGUF gemma-4-26B-A4B-GGUF-Q4_K_M.gguf --local-dir .(์๋-m๋ช ๋ น์ ํ์, ๋ค๋ฅธ ์์ํ๋ ํ์ผ๋ช ์Q4_K_M๋ง ๋ณ๊ฒฝ) - ์๋ฒ ์คํ:
llama-server -m gemma-4-26B-A4B-GGUF-Q4_K_M.gguf -c 8192 - ํ
์คํธ ์ด์ด์ฐ๊ธฐ:
llama-completion -m gemma-4-26B-A4B-GGUF-Q4_K_M.gguf -no-cnv -p "๋ํ๋ฏผ๊ตญ์ ์๋๋?" -n 128 - Ollama:
ollama run hf.co/axteam/gemma-4-26B-A4B-GGUF:Q4_K_M - KV ์บ์ ์ ๊ฐ(์ ํ):
--cache-type-k q8_0 --cache-type-v q8_0 -fa onโ KV ์บ์ ์ฝ ์ ๋ฐ, GGUF ํ์ผ๊ณผ ์๊ด์์ด ์คํ ์ ์ง์ - ๊ถ์ฅ ์ํ๋ง:
temperature=1.0,top_p=0.95,top_k=64(Google ๊ถ์ฅ๊ฐ)
๋ผ์ด์ ์ค
- ๋ผ์ด์ ์ค: Apache-2.0 (Gemma 4 ๋ผ์ด์ ์ค ์๋ด) โ ์๋ณธ ๋ชจ๋ธ์ ๋ผ์ด์ ์ค๋ฅผ ๊ทธ๋๋ก ๋ฐ๋ฆ
- ๊ถ๋ฆฌ: ์๋ณธ ๊ฐ์ค์นยทํ ํฌ๋์ด์ ๋ Google DeepMind ์์ , ์ด ๋ฆฌํฌ๋ ์์ํ ํ์๋ฌผ