NVIDIA-Nemotron-Nano-9B-v2-FP8-dynamic

Model Overview

• Model Architecture: NemotronHForCausalLM
  • Input: Text
  • Output: Text
• Model Optimizations:
  • Weight quantization: FP8
  • Activation quantization: FP8
• Release Date: 9/30/2025
• Version: 1.0
• Model Developers: Red Hat

Quantized version of nvidia/NVIDIA-Nemotron-Nano-9B-v2.

Model Optimizations

This model was obtained by quantizing the weights and activations of nvidia/NVIDIA-Nemotron-Nano-9B-v2 to FP8 data type. This optimization reduces the number of bits per parameter from 16 to 8, reducing the disk size and GPU memory requirements by approximately 50%. Only the weights and activations of the linear operators within transformers blocks are quantized.

Deployment

Use with vLLM

1. Initialize vLLM server:

vllm serve RedHatAI/NVIDIA-Nemotron-Nano-9B-v2-FP8-dynamic

2. Send requests to the server:

from openai import OpenAI

# Modify OpenAI's API key and API base to use vLLM's API server.
openai_api_key = "EMPTY"
openai_api_base = "http://<your-server-host>:8000/v1"

client = OpenAI(
    api_key=openai_api_key,
    base_url=openai_api_base,
)

model = "RedHatAI/NVIDIA-Nemotron-Nano-9B-v2-FP8-dynamic"

messages = [
    {"role": "user", "content": "Give me a short introduction to large language model."},
]

outputs = client.chat.completions.create(
    model=model,
    messages=messages,
)

generated_text = outputs.choices[0].message.content
print(generated_text)

Deploy on Red Hat AI Inference Server

podman run --rm -it --device nvidia.com/gpu=all -p 8000:8000 \
 --ipc=host \
--env "HUGGING_FACE_HUB_TOKEN=$HF_TOKEN" \
--env "HF_HUB_OFFLINE=0" -v ~/.cache/vllm:/home/vllm/.cache \
--name=vllm \
registry.access.redhat.com/rhaiis/rh-vllm-cuda \
vllm serve \
--tensor-parallel-size 8 \
--max-model-len 32768  \
--enforce-eager --model RedHatAI/NVIDIA-Nemotron-Nano-9B-v2-FP8-dynamic

Deploy on Red Hat Openshift AI

# Setting up vllm server with ServingRuntime
# Save as: vllm-servingruntime.yaml
apiVersion: serving.kserve.io/v1alpha1
kind: ServingRuntime
metadata:
 name: vllm-cuda-runtime # OPTIONAL CHANGE: set a unique name
 annotations:
   openshift.io/display-name: vLLM NVIDIA GPU ServingRuntime for KServe
   opendatahub.io/recommended-accelerators: '["nvidia.com/gpu"]'
 labels:
   opendatahub.io/dashboard: 'true'
spec:
 annotations:
   prometheus.io/port: '8080'
   prometheus.io/path: '/metrics'
 multiModel: false
 supportedModelFormats:
   - autoSelect: true
     name: vLLM
 containers:
   - name: kserve-container
     image: quay.io/modh/vllm:rhoai-2.25-cuda # CHANGE if needed. If AMD: quay.io/modh/vllm:rhoai-2.25-rocm
     command:
       - python
       - -m
       - vllm.entrypoints.openai.api_server
     args:
       - "--port=8080"
       - "--model=/mnt/models"
       - "--served-model-name={{.Name}}"
     env:
       - name: HF_HOME
         value: /tmp/hf_home
     ports:
       - containerPort: 8080
         protocol: TCP

# Attach model to vllm server. This is an NVIDIA template
# Save as: inferenceservice.yaml
apiVersion: serving.kserve.io/v1beta1
kind: InferenceService
metadata:
  annotations:
    openshift.io/display-name: NVIDIA-Nemotron-Nano-9B-v2-FP8-dynamic # OPTIONAL CHANGE
    serving.kserve.io/deploymentMode: RawDeployment
  name: NVIDIA-Nemotron-Nano-9B-v2-FP8-dynamic         # specify model name. This value will be used to invoke the model in the payload
  labels:
    opendatahub.io/dashboard: 'true'
spec:
  predictor:
    maxReplicas: 1
    minReplicas: 1
    model:
      modelFormat:
        name: vLLM
      name: ''
      resources:
        limits:
          cpu: '2'			# this is model specific
          memory: 8Gi		# this is model specific
          nvidia.com/gpu: '1'	# this is accelerator specific
        requests:			# same comment for this block
          cpu: '1'
          memory: 4Gi
          nvidia.com/gpu: '1'
      runtime: vllm-cuda-runtime	# must match the ServingRuntime name above
      storageUri: oci://registry.redhat.io/rhelai1/modelcar-nvidia-nemotron-nano-9b-v2-fp8-dynamic:1.5
    tolerations:
    - effect: NoSchedule
      key: nvidia.com/gpu
      operator: Exists

# make sure first to be in the project where you want to deploy the model
# oc project <project-name>

# apply both resources to run model

# Apply the ServingRuntime
oc apply -f vllm-servingruntime.yaml

# Replace <inference-service-name> and <cluster-ingress-domain> below:
# - Run `oc get inferenceservice` to find your URL if unsure.

# Call the server using curl:
curl https://<inference-service-name>-predictor-default.<domain>/v1/chat/completions
        -H "Content-Type: application/json" \
        -d '{
    "model": "NVIDIA-Nemotron-Nano-9B-v2-FP8-dynamic",
    "stream": true,
    "stream_options": {
        "include_usage": true
    },
    "max_tokens": 1,
    "messages": [
        {
            "role": "user",
            "content": "How can a bee fly when its wings are so small?"
        }
    ]
}'

See Red Hat Openshift AI documentation for more details.

Creation

This model was created with llm-compressor by running the code snippet below.

Model Creation Code

from llmcompressor.modifiers.quantization import QuantizationModifier
from llmcompressor.transformers import oneshot
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model
model_stub = "nvidia/NVIDIA-Nemotron-Nano-9B-v2"
model_name = model_stub.split("/")[-1]

model = AutoModelForCausalLM.from_pretrained(model_stub, dtype="auto")

tokenizer = AutoTokenizer.from_pretrained(model_stub)

# Configure the quantization algorithm and scheme
recipe = QuantizationModifier(
    ignore=["lm_head", "NemotronHMamba2Mixer"],
    targets="Linear",
    scheme="FP8_dynamic",
)

# Apply quantization
oneshot(
    model=model,
    recipe=recipe,
)

# Save to disk in compressed-tensors format
save_path = model_name + "-FP8-dynamic"
model.save_pretrained(save_path)
tokenizer.save_pretrained(save_path)
print(f"Model and tokenizer saved to: {save_path}")

Evaluation

The model was evaluated on the OpenLLMv1 leaderboard task, using lm-evaluation-harness, and on reasoning tasks using lighteval. vLLM was used for all evaluations.

Evaluation details

lm-evaluation-harness

lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/NVIDIA-Nemotron-Nano-9B-v2-FP8-dynamic",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1,gpu_memory_utilization=0.6,enable_chunked_prefill=True \
  --tasks openllm \
  --write_out \
  --batch_size auto \
  --output_path output_dir \
  --show_config

lighteval

lighteval_model_arguments.yaml

model_parameters:
  model_name: RedHatAI/NVIDIA-Nemotron-Nano-9B-v2-FP8-dynamic
  dtype: auto
  system_prompt: /think
  gpu_memory_utilization: 0.9
  generation_parameters:
    temperature: 0.6
    min_p: 0.0
    top_p: 0.95
    max_new_tokens: 32768

lighteval vllm \
  --model_args lighteval_model_arguments.yaml \
  --tasks lighteval|aime25|0 \

lighteval vllm \
  --model_args lighteval_model_arguments.yaml \
  --tasks lighteval|math_500|0 \

lighteval vllm \
  --model_args lighteval_model_arguments.yaml \
  --tasks lighteval|gpqa:diamond|0 \
  --use_chat_template = true

Accuracy

Category	Metric	nvidia/NVIDIA-Nemotron-Nano-9B-v2	RedHatAI/NVIDIA-Nemotron-Nano-9B-v2-FP8-dynamic	Recovery (%)
OpenLLM V1	ARC-Challenge (Acc-Norm, 25-shot)	64.16	63.23	98.5
	GSM8K (Strict-Match, 5-shot)	85.90	86.50	100.7
	HellaSwag (Acc-Norm, 10-shot)	79.57	79.75	100.2
	MMLU (Acc, 5-shot)	74.66	74.51	99.8
	TruthfulQA (MC2, 0-shot)	56.90	55.90	98.2
	Winogrande (Acc, 5-shot)	75.61	75.61	100.0
	Average Score	72.80	72.58	99.7
Reasoning (generation)	AIME 2025*	56.67	53.33	94.1
	GPQA diamond*	55.05	56.06	101.8
	Math-500*	95.90	95.47	99.6

* Average over 8 executions