Instructions to use Edge0/Edge0-120B-NVFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use Edge0/Edge0-120B-NVFP4 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 Edge0/Edge0-120B-NVFP4:BF16 # Run inference directly in the terminal: llama cli -hf Edge0/Edge0-120B-NVFP4:BF16
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf Edge0/Edge0-120B-NVFP4:BF16 # Run inference directly in the terminal: llama cli -hf Edge0/Edge0-120B-NVFP4:BF16
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 Edge0/Edge0-120B-NVFP4:BF16 # Run inference directly in the terminal: ./build/bin/llama cli -hf Edge0/Edge0-120B-NVFP4:BF16
Use Docker
docker model run hf.co/Edge0/Edge0-120B-NVFP4:BF16
- Llama new
- LM Studio
- Jan
- vLLM
How to use Edge0/Edge0-120B-NVFP4 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Edge0/Edge0-120B-NVFP4" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Edge0/Edge0-120B-NVFP4", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Edge0/Edge0-120B-NVFP4:BF16
- Ollama
How to use Edge0/Edge0-120B-NVFP4 with Ollama:
ollama run hf.co/Edge0/Edge0-120B-NVFP4:BF16
- Unsloth Desktop
- Pi
How to use Edge0/Edge0-120B-NVFP4 with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf Edge0/Edge0-120B-NVFP4:BF16
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:9931/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "Edge0/Edge0-120B-NVFP4:BF16" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use Edge0/Edge0-120B-NVFP4 with Docker Model Runner:
docker model run hf.co/Edge0/Edge0-120B-NVFP4:BF16
- Lemonade
How to use Edge0/Edge0-120B-NVFP4 with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull Edge0/Edge0-120B-NVFP4:BF16
Run and chat with the model
lemonade run user.Edge0-120B-NVFP4-BF16
List all available models
lemonade list
- Hermes Agent
How to use Edge0/Edge0-120B-NVFP4 with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf Edge0/Edge0-120B-NVFP4:BF16
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://127.0.0.1:9931/v1 hermes config set model.default Edge0/Edge0-120B-NVFP4:BF16
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use Edge0/Edge0-120B-NVFP4 with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf Edge0/Edge0-120B-NVFP4:BF16
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://127.0.0.1:9931/v1 \ --custom-model-id "Edge0/Edge0-120B-NVFP4:BF16" \ --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"
latent_dense_q8.bin / latent_dense_q4.bin re-uploaded — the previous D/U pair misbehaved in agent workloads
We have replaced the latent D/U projections in both 120B repositories. If you pulled the pack before
today, please re-download the two files below.
What was wrong. The previous latent_dense_q8.bin / latent_dense_q4.bin pair had a defect that
only showed up in agent workloads — long multi-turn sessions with tool calls. Short chat looked fine,
which is why it went unnoticed at release. The routed experts were correct; the problem was in the
projections that map between the hidden space and the latent space.
What changed. Both files are rebuilt from the current agent-training step (step 6000). Nothing else
in the pack changed: same format, same size, same experts.bin, same skeleton and MTP. It is a drop-in
replacement, no config change.
| file | size | sha256 |
|---|---|---|
latent_dense_q8.bin |
2,673,868,800 | 1e82f0e02545734f… |
latent_dense_q4.bin |
1,415,577,600 | 9ec1974b83bd4bcf… |
Both repositories (Edge0-120B-NVFP4 and Edge0-120B-q3k) now carry the same new pair.
How to update. The Hub keeps the same file names, so a cached copy will not refresh by itself —
delete the two files first, then download:
rm -f latent_dense_q8.bin latent_dense_q4.bin
huggingface-cli download Edge0/Edge0-120B-NVFP4 --include "latent_dense_*" --local-dir .