Instructions to use luxopes/LuxAI-1-Nemo-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 luxopes/LuxAI-1-Nemo-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 luxopes/LuxAI-1-Nemo-GGUF:F16
# Run inference directly in the terminal:
llama cli -hf luxopes/LuxAI-1-Nemo-GGUF:F16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf luxopes/LuxAI-1-Nemo-GGUF:F16
# Run inference directly in the terminal:
llama cli -hf luxopes/LuxAI-1-Nemo-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 luxopes/LuxAI-1-Nemo-GGUF:F16
# Run inference directly in the terminal:
./llama-cli -hf luxopes/LuxAI-1-Nemo-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 luxopes/LuxAI-1-Nemo-GGUF:F16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf luxopes/LuxAI-1-Nemo-GGUF:F16

Use Docker

docker model run hf.co/luxopes/LuxAI-1-Nemo-GGUF:F16

LM Studio
Jan

vLLM

How to use luxopes/LuxAI-1-Nemo-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "luxopes/LuxAI-1-Nemo-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": "luxopes/LuxAI-1-Nemo-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/luxopes/LuxAI-1-Nemo-GGUF:F16

Ollama
How to use luxopes/LuxAI-1-Nemo-GGUF with Ollama:
```
ollama run hf.co/luxopes/LuxAI-1-Nemo-GGUF:F16
```
Unsloth Desktop

How to use luxopes/LuxAI-1-Nemo-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf luxopes/LuxAI-1-Nemo-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": "luxopes/LuxAI-1-Nemo-GGUF:F16"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use luxopes/LuxAI-1-Nemo-GGUF with Docker Model Runner:
```
docker model run hf.co/luxopes/LuxAI-1-Nemo-GGUF:F16
```

Lemonade

How to use luxopes/LuxAI-1-Nemo-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull luxopes/LuxAI-1-Nemo-GGUF:F16

Run and chat with the model

lemonade run user.LuxAI-1-Nemo-GGUF-F16

List all available models

lemonade list

Hermes Agent

How to use luxopes/LuxAI-1-Nemo-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 luxopes/LuxAI-1-Nemo-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 luxopes/LuxAI-1-Nemo-GGUF:F16

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use luxopes/LuxAI-1-Nemo-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf luxopes/LuxAI-1-Nemo-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 "luxopes/LuxAI-1-Nemo-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"

LuxAI 1 Nemo — GGUF

Complete exports of the corrected Nemo model. See the main model card for training, usage, evaluations, and limitations.

File	Decimal GB	Description
LuxAI-1-Nemo-F16.gguf	27.021	F16 converted from merged BF16
LuxAI-1-Nemo-Q4_K_M.gguf	9.996	Q4_K_M, mixed 4-bit quantization
mmproj-LuxAI-1-Nemo-F16.gguf	0.878	Frozen vision projector

Format	Natural valid tool calls	Native thinking in tool probes	Complete standalone native thinking
F16	8/8	8/8	2/4
Q4_K_M	8/8	8/8	2/4

The Q4_K_M export keeps token embeddings and the output matrix in F16, byte-identical to the F16 export. Transformer weights use mixed 4-bit quantization. This preserves the trained token rows and makes the file larger than a standard Q4_K_M export.

Both text exports passed actual GPU generation and OpenAI-style tool API tests, including a follow-up answer after a simulated tool result. All 240 Czech Lux benchmark prompts passed tokenization parity checks. These focused checks are separate from the merged BF16 Lux score of 72.92%; a GGUF Lux accuracy score is not claimed.

Natural GGUF tool probes: F16: 8/8, Q4_K_M: 8/8. Raw generation and the chat API agreed on tool presence. The API checks separately establish parsing and a successful call/result roundtrip; they do not guarantee that the model will choose a tool whenever requested.

Standalone thinking used the same four prompts and 2,048-token output limit as the merged BF16 probe. Completion requires native opening/closing tokens and EOS.

Use the embedded native chat template and a runtime with mistral3 and native V13 tool support. Actual thinking tokens are [THINK] / [/THINK] (34/35), and tool syntax is [TOOL_CALLS]function_name[ARGS]{"argument":"value"} (9/32). Start with an 8,192-token context. Reasoning is not guaranteed on every prompt, and long output can be truncated by the client. Vision quality was not tested.

The tested server runtime was llama.cpp b10830 (commit 465e49b9cea78a68b9c244ffb48d0ee24a82873d). A corresponding launch command is:

llama-server -m LuxAI-1-Nemo-Q4_K_M.gguf -ngl 99 -c 8192 --jinja --reasoning-format none

With this tested setting, thinking delimiters remain in the API message content; the client separates [THINK]...[/THINK] for display. Native function calls are returned as tool_calls. Do not assume every client recognizes the reasoning format automatically. The hosted ZeroGPU chat includes its own reasoning panel.

Exact source revision, separate converter/runtime versions, hashes, and test receipts are in GGUF_MANIFEST.json. Component notices and data attribution are preserved in the accompanying files.

Downloads last month: -

GGUF

Model size

14B params

Architecture

mistral3

Hardware compatibility

4-bit

16-bit

Model tree for luxopes/LuxAI-1-Nemo-GGUF

Base model

luxopes/LuxAI-1-Nemo

Quantized

(1)

this model