Instructions to use LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4 with Transformers:

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

pipe = pipeline("text-generation", model="LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4", trust_remote_code=True, 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

vLLM

How to use LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-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": "LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4

SGLang

How to use LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4 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 "LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4" \
    --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": "LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4",
		"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 "LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4" \
        --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": "LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4 with Docker Model Runner:
```
docker model run hf.co/LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4
```

NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4

A coding-specialized expansion of NVIDIA Nemotron 3 Super 120B-A12B NVFP4.

This is a community model and is not an official NVIDIA release. It was inspired by NVIDIA's open-model, open-data, and open-tooling work around Nemotron.

Model Summary


Total Parameters	Approximately 145B
Active Parameters	Approximately 13B per token
Quantization	NVFP4 mixed-precision checkpoint
Architecture	Nemotron hybrid Mamba-2, LatentMoE, Attention, and MTP
Maximum Context Length	262,144 tokens (256K)
Specialization	Agentic coding, reasoning, tool use, and multi-turn software-engineering workflows
Base Model	NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4

Does This Work?

The public Nemotron 130B LLMWildling Canary NVFP4 provides a smaller proof point. It demonstrates direct recall of newly added domain knowledge and carries that knowledge into a follow-up task without RAG or prompt-injected context.

SVG Evaluation

This checkpoint received substantial agentic software-engineering training. During interactive evaluation, it also produced these playable SVG Flappy Bird-style outputs.

Evaluation 1	Evaluation 2

Intended Use

This checkpoint is ready for production use in coding assistants, repository analysis, agentic software-engineering systems, tool-using workflows, and interactive SVG generation.

License

This model is derived from NVIDIA Nemotron 3 Super. Use is governed by the NVIDIA Nemotron Open Model License. Review the upstream model card for its full terms, safety information, limitations, and base-model details.

Downloads last month: 383

Safetensors

Model size

88B params

Tensor type

F32

BF16

F8_E4M3

Model tree for LLMWildling/NVIDIA-Nemotron-3-Super-145B-A13B-Coder-NVFP4

Base model

nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4

Quantized

(11)

this model