Instructions to use Akira-Papa/akira-papa-1.0-31b-jp with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Akira-Papa/akira-papa-1.0-31b-jp with MLX:

# Download the model from the Hub
pip install huggingface_hub[hf_xet]

huggingface-cli download --local-dir akira-papa-1.0-31b-jp Akira-Papa/akira-papa-1.0-31b-jp

llama-cpp-python

How to use Akira-Papa/akira-papa-1.0-31b-jp with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Akira-Papa/akira-papa-1.0-31b-jp",
	filename="akira-papa-1.0-31b-jp-Q4_K_M.gguf",
)

llm.create_chat_completion(
	messages = "No input example has been defined for this model task."
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use Akira-Papa/akira-papa-1.0-31b-jp with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Akira-Papa/akira-papa-1.0-31b-jp:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Akira-Papa/akira-papa-1.0-31b-jp:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Akira-Papa/akira-papa-1.0-31b-jp:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Akira-Papa/akira-papa-1.0-31b-jp:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf Akira-Papa/akira-papa-1.0-31b-jp:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf Akira-Papa/akira-papa-1.0-31b-jp:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf Akira-Papa/akira-papa-1.0-31b-jp:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Akira-Papa/akira-papa-1.0-31b-jp:Q4_K_M

Use Docker

docker model run hf.co/Akira-Papa/akira-papa-1.0-31b-jp:Q4_K_M

LM Studio
Jan
Ollama
How to use Akira-Papa/akira-papa-1.0-31b-jp with Ollama:
```
ollama run hf.co/Akira-Papa/akira-papa-1.0-31b-jp:Q4_K_M
```

Unsloth Studio

How to use Akira-Papa/akira-papa-1.0-31b-jp with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Akira-Papa/akira-papa-1.0-31b-jp to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Akira-Papa/akira-papa-1.0-31b-jp to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Akira-Papa/akira-papa-1.0-31b-jp to start chatting

How to use Akira-Papa/akira-papa-1.0-31b-jp with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "Akira-Papa/akira-papa-1.0-31b-jp"

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "Akira-Papa/akira-papa-1.0-31b-jp"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use Akira-Papa/akira-papa-1.0-31b-jp with Hermes Agent:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "Akira-Papa/akira-papa-1.0-31b-jp"

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default Akira-Papa/akira-papa-1.0-31b-jp

Run Hermes

hermes

Docker Model Runner
How to use Akira-Papa/akira-papa-1.0-31b-jp with Docker Model Runner:
```
docker model run hf.co/Akira-Papa/akira-papa-1.0-31b-jp:Q4_K_M
```

Lemonade

How to use Akira-Papa/akira-papa-1.0-31b-jp with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Akira-Papa/akira-papa-1.0-31b-jp:Q4_K_M

Run and chat with the model

lemonade run user.akira-papa-1.0-31b-jp-Q4_K_M

List all available models

lemonade list

akira-papa-1.0-31b-jp

akira-papa-1.0-31b-jp は、Gemma 4 31B-it を土台に、Akira-Papa が作成・整理した日本語データ、rewrite、route-aware tuning を重ねた 31B flagship release です。
自然な日本語の長文、ブログ / note の本文、レビュー、初学者向けの技術説明、軽中量のコーディング支援を 1 本で扱いやすくすることを主目的に整えています。

このリポジトリには、base model と adapter を結合した fused model に加えて、GGUF 版の Q8_0 / Q4_K_M も含みます。

位置づけ

現在の系統: 31B flagship
Hugging Face repo: https://huggingface.co/Akira-Papa/akira-papa-1.0-31b-jp
Gemma 4 31B-it ベースの日本語 flagship ライン
blog / note / wall-bounce / coding support のバランス版

主な用途

日本語のブログ本文
note 風の自然な長文
実装方針やレビュー観点の整理
初学者向けのやさしい技術説明
軽中量のコーディング支援
壁打ち相手としての論点整理

強み

日本語の自然文が崩れにくい
長文 writer でも tone を保ちやすい
説明、レビュー、壁打ち、coding support を 1 本で両立しやすい
Q8_0 / Q4_K_M により llama.cpp / LM Studio / Windows 系 runtime にも載せやすい

評価の要約

現在の評価メモに記録している代表値は次です。

writer hard eval v2: 1.00 / 1.00
writer longform eval v3: 1.00 / 1.00
writer blog voice eval v4: 1.00 / 1.00
coding weak5: 2.80 / 3.00

同梱される主なファイル

fused model 一式
config.json
tokenizer.json
model-00001-of-00004.safetensors
model-00002-of-00004.safetensors
model-00003-of-00004.safetensors
model-00004-of-00004.safetensors
akira-papa-1.0-31b-jp-Q8_0.gguf
akira-papa-1.0-31b-jp-Q4_K_M.gguf

使い分け

fused model: Apple Silicon / MLX の folder 読み込み向け
Q8_0: GGUF 側の品質優先
Q4_K_M: GGUF 側の容量優先
31B は依然として重いので、軽量配布が必要なら E2B / E4B line も検討してください

目安サイズ:

fused model directory 全体で約 16 GB
Q8_0 は約 30 GB
Q4_K_M は約 17 GB

注意

coding 専用 best を最大化したモデルではありません
外部情報の正確性は別途検証が必要です
ローカル実行コストは重めです
Q8_0 / Q4_K_M は llama.cpp / LM Studio / Windows 系 runtime 向けです
fused model は Apple Silicon / MLX 前提の運用が最も扱いやすいです

ベースモデルと利用条件

このモデルのベースは google/gemma-4-31b-it です。

利用前に次を確認してください。

NOTICE
GEMMA_TERMS.md
MODIFICATIONS.md

Downloads last month: 267

Safetensors

Model size

31B params

Tensor type

BF16

U32

MLX

Hardware compatibility

4-bit

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Akira-Papa
/

akira-papa-1.0-31b-jp