Instructions to use Ginckl/HashBlitz-1-7B-Max with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Ginckl/HashBlitz-1-7B-Max with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Ginckl/HashBlitz-1-7B-Max")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Ginckl/HashBlitz-1-7B-Max")
model = AutoModelForCausalLM.from_pretrained("Ginckl/HashBlitz-1-7B-Max", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use Ginckl/HashBlitz-1-7B-Max 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 Ginckl/HashBlitz-1-7B-Max:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Ginckl/HashBlitz-1-7B-Max:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Ginckl/HashBlitz-1-7B-Max:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Ginckl/HashBlitz-1-7B-Max: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 Ginckl/HashBlitz-1-7B-Max:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf Ginckl/HashBlitz-1-7B-Max: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 Ginckl/HashBlitz-1-7B-Max:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Ginckl/HashBlitz-1-7B-Max:Q4_K_M

Use Docker

docker model run hf.co/Ginckl/HashBlitz-1-7B-Max:Q4_K_M

LM Studio
Jan

vLLM

How to use Ginckl/HashBlitz-1-7B-Max with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Ginckl/HashBlitz-1-7B-Max"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Ginckl/HashBlitz-1-7B-Max",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Ginckl/HashBlitz-1-7B-Max:Q4_K_M

SGLang

How to use Ginckl/HashBlitz-1-7B-Max with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Ginckl/HashBlitz-1-7B-Max" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Ginckl/HashBlitz-1-7B-Max",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Ginckl/HashBlitz-1-7B-Max" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Ginckl/HashBlitz-1-7B-Max",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use Ginckl/HashBlitz-1-7B-Max with Ollama:
```
ollama run hf.co/Ginckl/HashBlitz-1-7B-Max:Q4_K_M
```
Unsloth Desktop

How to use Ginckl/HashBlitz-1-7B-Max with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Ginckl/HashBlitz-1-7B-Max: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": "Ginckl/HashBlitz-1-7B-Max:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use Ginckl/HashBlitz-1-7B-Max with Docker Model Runner:
```
docker model run hf.co/Ginckl/HashBlitz-1-7B-Max:Q4_K_M
```

Lemonade

How to use Ginckl/HashBlitz-1-7B-Max with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Ginckl/HashBlitz-1-7B-Max:Q4_K_M

Run and chat with the model

lemonade run user.HashBlitz-1-7B-Max-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use Ginckl/HashBlitz-1-7B-Max with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Ginckl/HashBlitz-1-7B-Max: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 Ginckl/HashBlitz-1-7B-Max:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use Ginckl/HashBlitz-1-7B-Max with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Ginckl/HashBlitz-1-7B-Max: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 "Ginckl/HashBlitz-1-7B-Max: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"

HashBlitz 1 Max 7B

Originally conceived as a strictly private build for internal deployment, HashBlitz 1 Max 7B was never intended to see a public release. However, the remarkable synergy and performance achieved during our training phase compelled us to share these custom weights with the open-source community, honoring the best practices of AI transparency.

This project transcends a simple experimental fine-tune of the Qwen 2.5 architecture. Instead, it represents a deliberately crafted amalgamation of top-tier artificial intelligence, distilled into a highly optimized 7-billion parameter foundation. We built these custom weights by synthesizing specialized, meticulously curated intelligence from the industry's most advanced frontier models.

To create a deeply capable and well-rounded system, we isolated and extracted the absolute best traits from four distinct intelligence paradigms:

Claude Opus 5 forged the model's core coding intelligence, ensuring profound logical reasoning and structural elegance in complex software development tasks.
Groq 4.5 provided the foundational character and safety alignment, ensuring that the model's behavior remains highly engaging, nuanced, and strictly responsible.
GPT-5.4 shaped the everyday conversational approach, bringing a natural, fluid warmth and high emotional intelligence to continuous dialogue.
Gemini 3.1 Pro (Extended Thinking) grounded the model in real-world complexity, capturing the cognitive behaviors required for processing live, updated information and deep-dive analytical tasks.

By releasing these custom weights, our intended use is to empower developers and vibe coders with a profoundly capable local agent that punches far above its weight class. We invite you to explore this unique convergence of cutting-edge reasoning paradigms, now fully open and available.