Model Intelligence Sheet
JMingo/gemma-4-26B-A4B-it-qat-japanese-pruned-GGUF overview
Expert pruned variants of gemma 4 26B A4B it qat q4 0 unquantized https://huggingface.co/google/gemma 4 26B A4B it qat q4 0 unquantized Experts were selected a…
Runs locally from ~4.27 GB disk (8 GB VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
13 GGUF files detected
Direct downloads for local inference
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| gemma-4-26B-A4B-it-qat-japanese-104e-Q4_0.gguf | GGUF | Q4_0 | 11.02 GB | Download |
| gemma-4-26B-A4B-it-qat-japanese-112e-Q4_0.gguf | GGUF | Q4_0 | 11.77 GB | Download |
| gemma-4-26B-A4B-it-qat-japanese-120e-Q4_0.gguf | GGUF | Q4_0 | 12.52 GB | Download |
| gemma-4-26B-A4B-it-qat-japanese-128e-Q4_0.gguf | GGUF | Q4_0 | 13.27 GB | Download |
| gemma-4-26B-A4B-it-qat-japanese-32e-Q4_0.gguf | GGUF | Q4_0 | 4.27 GB | Download |
| gemma-4-26B-A4B-it-qat-japanese-40e-Q4_0.gguf | GGUF | Q4_0 | 5.02 GB | Download |
| gemma-4-26B-A4B-it-qat-japanese-48e-Q4_0.gguf | GGUF | Q4_0 | 5.77 GB | Download |
| gemma-4-26B-A4B-it-qat-japanese-56e-Q4_0.gguf | GGUF | Q4_0 | 6.52 GB | Download |
| gemma-4-26B-A4B-it-qat-japanese-64e-Q4_0.gguf | GGUF | Q4_0 | 7.27 GB | Download |
| gemma-4-26B-A4B-it-qat-japanese-72e-Q4_0.gguf | GGUF | Q4_0 | 8.02 GB | Download |
| gemma-4-26B-A4B-it-qat-japanese-80e-Q4_0.gguf | GGUF | Q4_0 | 8.77 GB | Download |
| gemma-4-26B-A4B-it-qat-japanese-88e-Q4_0.gguf | GGUF | Q4_0 | 9.52 GB | Download |
| gemma-4-26B-A4B-it-qat-japanese-96e-Q4_0.gguf | GGUF | Q4_0 | 10.27 GB | Download |
Model Details
Model README
---
license: apache-2.0
base_model:
- JMingo/gemma-4-26B-A4B-it-qat-GGUF
tags:
- gemma4
---
Expert-pruned variants of gemma-4-26B-A4B-it-qat-q4_0-unquantized
Experts were selected and pruned via a Japanese imatrix to preserve Japanese language proficiency.
Model quality remains stable with 72 or more experts retained.
---
Benchmarks
Run JMingo/gemma-4-26B-A4B-it-qat-japanese-pruned-GGUF with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models