Instructions to use OsGo/Qwen3.5-ocr-jp-2b-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use OsGo/Qwen3.5-ocr-jp-2b-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16
# Run inference directly in the terminal:
llama cli -hf OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16
# Run inference directly in the terminal:
llama cli -hf OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16

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 OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16
# Run inference directly in the terminal:
./llama-cli -hf OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16

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 OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16

Use Docker

docker model run hf.co/OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16

LM Studio
Jan

vLLM

How to use OsGo/Qwen3.5-ocr-jp-2b-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "OsGo/Qwen3.5-ocr-jp-2b-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "OsGo/Qwen3.5-ocr-jp-2b-GGUF",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16

Ollama
How to use OsGo/Qwen3.5-ocr-jp-2b-GGUF with Ollama:
```
ollama run hf.co/OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16
```
Unsloth Desktop

How to use OsGo/Qwen3.5-ocr-jp-2b-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use OsGo/Qwen3.5-ocr-jp-2b-GGUF with Docker Model Runner:
```
docker model run hf.co/OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16
```

Lemonade

How to use OsGo/Qwen3.5-ocr-jp-2b-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16

Run and chat with the model

lemonade run user.Qwen3.5-ocr-jp-2b-GGUF-F16

List all available models

lemonade list

Hermes Agent

How to use OsGo/Qwen3.5-ocr-jp-2b-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16

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 OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use OsGo/Qwen3.5-ocr-jp-2b-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "OsGo/Qwen3.5-ocr-jp-2b-GGUF:F16" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

⚠️ Mirror / Backup Notice: This repository is an unmodified backup.

Original GGUF Repo: nakanishi08/Qwen3.5-ocr-jp-2b-GGUF

Base Model: ebinan92/Qwen3.5-ocr-jp-2b

For the latest updates, issues, and support, please visit the upstream repository.

Qwen3.5-OCR-JP-2B — GGUF

GGUF quantized versions of ebinan92/Qwen3.5-ocr-jp-2b, a Japanese/English OCR vision-language model based on Qwen3.5-2B.

Converted and quantized using llama.cpp b10488.

Files

File	Size	Description
`qwen35ocr-jp-2b-q4_k_m.gguf`	1.5 GB	Q4_K_M quantized (recommended)
`qwen35ocr-jp-2b-mmproj-f16.gguf`	638 MB	Multimodal projector — required for both files below
`qwen35ocr-jp-2b-f16.gguf`	4.6 GB	F16 full precision (for re-quantization)

Both qwen35ocr-jp-2b-q4_k_m.gguf and qwen35ocr-jp-2b-mmproj-f16.gguf are required to run inference.

Usage

llama-server

llama-server \
  -m qwen35ocr-jp-2b-q4_k_m.gguf \
  --mmproj qwen35ocr-jp-2b-mmproj-f16.gguf \
  -ngl 99 \
  --image-min-tokens 1024 \
  --host 127.0.0.1 --port 11436 \
  -c 8192 -n 4096

--host and --port can be changed to suit your environment. Use --host 0.0.0.0 to allow access from other machines on your network.

Ollama (Modelfile)

FROM ./qwen35ocr-jp-2b-q4_k_m.gguf
FROM ./qwen35ocr-jp-2b-mmproj-f16.gguf

ollama create qwen35ocr-jp-2b -f Modelfile

Performance

Measured with Q4_K_M + llama-server b10488 on a single Japanese document page (A4 scanned PDF).

GPU	VRAM	Speed	tok/s
NVIDIA RTX A2000	12 GB	~15 sec/page	~98 tok/s
NVIDIA Quadro P2000	4 GB	~100 sec/page	~43 tok/s

Conversion

# 1. Convert HF model to F16 GGUF
python convert_hf_to_gguf.py ebinan92/Qwen3.5-ocr-jp-2b --outtype f16

# 2. Quantize to Q4_K_M
llama-quantize qwen35ocr-jp-2b-f16.gguf qwen35ocr-jp-2b-q4_k_m.gguf Q4_K_M

Quantization took ~47 seconds on RTX A2000.

About the Original Model

ebinan92/Qwen3.5-ocr-jp-2b is a Japanese/English OCR model fine-tuned from Qwen/Qwen3.5-2B-Base. Key features:

Output format: Chandra OCR 2 compatible HTML layout blocks with bounding boxes and semantic labels
Japanese handwriting support
Vertical text support
Ruby (furigana) annotation support (HTML5 <ruby> markup)

Acknowledgements

ebinan92 for creating and sharing Qwen3.5-OCR-JP-2B
Qwen Team for the base model Qwen3.5-2B
SandLogic Technologies for Chandra OCR 2, whose output format this model is compatible with (no weights or outputs from Chandra were used)

License

Apache 2.0 — same as the original model.

Downloads last month: 289

GGUF

Model size

2B params

Architecture

qwen35

Hardware compatibility

4-bit

16-bit

Model tree for OsGo/Qwen3.5-ocr-jp-2b-GGUF

Base model

Qwen/Qwen3.5-2B-Base

Finetuned

Qwen/Qwen3.5-2B

Finetuned

ebinan92/Qwen3.5-ocr-jp-2b

Quantized

(2)

this model