Instructions to use wickgraveyard/qwen35b-nvfp4-blackwell with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use wickgraveyard/qwen35b-nvfp4-blackwell with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="wickgraveyard/qwen35b-nvfp4-blackwell",
	filename="qwen35b-nvfp4-blackwell.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 wickgraveyard/qwen35b-nvfp4-blackwell 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 wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4
# Run inference directly in the terminal:
llama cli -hf wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4
# Run inference directly in the terminal:
llama cli -hf wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4

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 wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4
# Run inference directly in the terminal:
./llama-cli -hf wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4

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 wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4
# Run inference directly in the terminal:
./build/bin/llama-cli -hf wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4

Use Docker

docker model run hf.co/wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4

LM Studio
Jan
Ollama
How to use wickgraveyard/qwen35b-nvfp4-blackwell with Ollama:
```
ollama run hf.co/wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4
```

Unsloth Studio

How to use wickgraveyard/qwen35b-nvfp4-blackwell 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 wickgraveyard/qwen35b-nvfp4-blackwell 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 wickgraveyard/qwen35b-nvfp4-blackwell to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for wickgraveyard/qwen35b-nvfp4-blackwell to start chatting

How to use wickgraveyard/qwen35b-nvfp4-blackwell with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use wickgraveyard/qwen35b-nvfp4-blackwell with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4

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 wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use wickgraveyard/qwen35b-nvfp4-blackwell with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4

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 "wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4" \
  --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"

Docker Model Runner
How to use wickgraveyard/qwen35b-nvfp4-blackwell with Docker Model Runner:
```
docker model run hf.co/wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4
```

Lemonade

How to use wickgraveyard/qwen35b-nvfp4-blackwell with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull wickgraveyard/qwen35b-nvfp4-blackwell:NVFP4

Run and chat with the model

lemonade run user.qwen35b-nvfp4-blackwell-NVFP4

List all available models

lemonade list

Qwen3.6-35B-Abliterated-Claude-4.7-200k — NVFP4 Blackwell GGUF

Architecture-specific format for NVIDIA Blackwell GPUs (sm_120/sm_121).

What is NVFP4?

NVFP4 is a native FP4 quantization format that engages the Blackwell FP4 tensor cores. On a DGX Spark / GB10, this delivers ~23% faster generation compared to Q4_K_M GGUF.

Note: This model has the MTP (Multi-Token Prediction) layer removed to work around a qwen35moe handler bug in Ollama v0.30.11. Normal single-token generation is unaffected.

Performance (GB10 DGX Spark)

Metric	Q4_K_M	NVFP4	Improvement
Generation	71.42 tok/s	87.65 tok/s	+23%
File size	23 GB	19 GB	-17%

Requirements

Hardware: NVIDIA Blackwell GPU (GB10, RTX 5090, etc.) — sm_120 or sm_121
Software: Ollama v0.30.11+ (CUDA v13 backend)
Other GPUs: Will fall back to CPU or standard CUDA — not recommended

Usage

ollama pull wickgraveyard/qwen35b-nvfp4-blackwell
ollama run wickgraveyard/qwen35b-nvfp4-blackwell

Conversion Method

This model was converted from Q4_K_M GGUF using llama-quantize with --prune-layers 40 (to remove the MTP layer) and --tensor-type-file for NVFP4 mapping.