Instructions to use prithivMLmods/inclusionAI-ZwZ2-9B-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 prithivMLmods/inclusionAI-ZwZ2-9B-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 prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama cli -hf prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M

Use Docker

docker model run hf.co/prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M

Llama new
LM Studio
Jan

vLLM

How to use prithivMLmods/inclusionAI-ZwZ2-9B-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "prithivMLmods/inclusionAI-ZwZ2-9B-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": "prithivMLmods/inclusionAI-ZwZ2-9B-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/prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M

Ollama
How to use prithivMLmods/inclusionAI-ZwZ2-9B-GGUF with Ollama:
```
ollama run hf.co/prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M
```
Unsloth Desktop

How to use prithivMLmods/inclusionAI-ZwZ2-9B-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M

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": "prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use prithivMLmods/inclusionAI-ZwZ2-9B-GGUF with Docker Model Runner:
```
docker model run hf.co/prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M
```

Lemonade

How to use prithivMLmods/inclusionAI-ZwZ2-9B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.inclusionAI-ZwZ2-9B-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use prithivMLmods/inclusionAI-ZwZ2-9B-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 prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M

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 prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use prithivMLmods/inclusionAI-ZwZ2-9B-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M

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 "prithivMLmods/inclusionAI-ZwZ2-9B-GGUF:Q4_K_M" \
  --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"

inclusionAI-ZwZ2-9B-GGUF

ZwZ2-9B (inclusionAI/ZwZ2-9B) is an open-weight 9B-parameter vision-language model from inclusionAI’s Zooming-without-Zooming-v2 collection, built on Qwen/Qwen3.5-9B to connect evidence across image regions through comparison, matching, aggregation, and multi-step reasoning without inference-time zooming tools. Trained directly with Group Relative Policy Optimization (GRPO), without an additional supervised fine-tuning stage, it uses approximately 20,000 examples from inclusionAI/ZwZ2-RL-VQA, generated through the EvidenceConnect pipeline on SA-1B images. The pipeline creates multi-region questions from 3–7 related regions, using privileged crops for data construction while training on full images with highlighted regions. Across 25 benchmarks spanning fine-grained perception, spatial reasoning, long-context visual understanding, multimodal agents, and general visual reasoning, ZwZ2-9B outperforms its Qwen3.5-9B base on all 25 benchmarks, improving equal-weighted accuracy from 51.63% to 56.59% (+4.96 points). Evaluation includes the 25 non-counting tasks of the MMLongBench* visual core, covering V-RAG, Image MM-NIAH, and VH-Multi. The model supports BF16 inference through Hugging Face Transformers (transformers==5.13.1) using Qwen3_5ForConditionalGeneration, AutoProcessor, and the included chat_template.jinja. It uses non-thinking mode and greedy decoding by default, wrapping final responses in answer tags unless overridden by a custom system message.

MTP — Speculative Decoding: No

Model Files

File Name	Quant Type	File Size	File Link	Description
ZwZ2-9B.BF16.gguf	BF16	17.9 GB	Link	Full BF16 weights. Highest quality, largest file size.
ZwZ2-9B.Q3_K_L.gguf	Q3_K_L	4.93 GB	Link	Lower quality but usable, good for low RAM availability.
ZwZ2-9B.Q3_K_M.gguf	Q3_K_M	4.62 GB	Link	Low quality.
ZwZ2-9B.Q4_K_M.gguf	Q4_K_M	5.63 GB	Link	Good quality, default size for most use cases, recommended.
ZwZ2-9B.Q4_K_S.gguf	Q4_K_S	5.35 GB	Link	Slightly lower quality with more space savings, recommended.
ZwZ2-9B.Q5_K_M.gguf	Q5_K_M	6.47 GB	Link	High quality, recommended.
ZwZ2-9B.Q5_K_S.gguf	Q5_K_S	6.31 GB	Link	High quality, recommended.
ZwZ2-9B.Q6_K.gguf	Q6_K	7.36 GB	Link	Very high quality, near perfect, recommended.
ZwZ2-9B.Q8_0.gguf	Q8_0	9.53 GB	Link	Extremely high quality, generally unneeded but max available quant.
ZwZ2-9B.mmproj-bf16.gguf	mmproj-bf16	922 MB	Link	Multimodal projection file in BF16 format. Used for vision/language models.