InferX Catalog | NVIDIA-Nemotron-3-Nano-30B-A3B-FP8

NVIDIA-Nemotron-3-Nano-30B-A3B-FP8

Optimized with FP8 for fast, efficient reasoning and chat workloads.

nvidia text text2text reasoning high-throughput agentic

Log in to deploy

nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8 is a 30B parameter Nemotron model from NVIDIA, optimized with FP8 quantization and A3B architecture for improved efficiency, strong reasoning, and high-throughput inference, making it well-suited for agents, coding, and production deployments on modern GPUs.

Metadata

Provider

nvidia

Modality

text

API type

text2text

Source

huggingface / nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8

Created

2026-03-31 16:35:12 UTC

Updated

2026-03-31 16:46:56 UTC

Catalog version

1

Visibility

Published

Specifications

Parameters

30.00B

MoE

No

Max model length

262144

Image

vllm/vllm-openai:v0.16.0

Default Deploy Config

GPU count

1

vRAM

70000 MB

Summary

1xGPU 70000 MB

Recommended Use Cases

Agent / tool use
Reasoning assistant

Model Spec

{
    "image": "vllm/vllm-openai:v0.16.0",
    "commands": [
        "--model",
        "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8",
        "--trust-remote-code",
        "--gpu-memory-utilization",
        "0.95",
        "--max-model-len",
        "262144",
        "--kv-cache-dtype",
        "fp8",
        "--max-num-seqs",
        "8",
        "--trust-remote-code",
        "--enable-auto-tool-choice",
        "--tool-call-parser",
        "qwen3_coder",
        "--reasoning-parser-plugin",
        "/root/.cache/huggingface/hub/models--nvidia--NVIDIA-Nemotron-3-Nano-30B-A3B-FP8/snapshots/f8dc1c0afee92f44417695b4f5ddca9afc95ea58/nano_v3_reasoning_parser.py",
        "--reasoning-parser",
        "nano_v3",
        "--kv-cache-dtype",
        "fp8",
        "--tensor-parallel-size=1"
    ],
    "resources": {
        "GPU": {
            "Count": 1,
            "vRam": 70000
        }
    },
    "envs": [],
    "policy": {
        "Obj": {
            "min_replica": 0,
            "max_replica": 1,
            "standby_per_node": 1,
            "parallel": 50,
            "queue_len": 100,
            "queue_timeout": 30.0,
            "scalein_timeout": 1.0,
            "scaleout_policy": {
                "WaitQueueRatio": {
                    "wait_ratio": 0.1
                }
            },
            "runtime_config": {
                "graph_sync": false
            }
        }
    },
    "sample_query": {
        "body": {
            "model": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8",
            "stream": "true",
            "max_tokens": "1000",
            "temperature": "0"
        },
        "path": "v1/completions",
        "prompt": "write a quick sort algorithm.",
        "apiType": "text2text",
        "dataUrl": "",
        "prompts": [
            "Write a Python function that computes Fibonacci numbers. Explain time complexity.",
            "Translate the following Chinese text to English: \u4eca\u5929\u5929\u6c14\u5f88\u597d\u3002",
            "Explain general relativity in simple language.",
            "Write a legal contract clause about liability and indemnification.",
            "Summarize the plot of a fantasy novel involving dragons.",
            "Solve this calculus integral: \u222b x^3 log(x) dx",
            "Generate a JSON schema describing a user profile.",
            "Explain why emojis like \ud83d\ude00\ud83d\udd25\ud83d\ude80 represent byte-level tokens."
        ],
        "loadingTimeout": 90
    }
}