benchmarks / nemotron-3-nano-30b-a3b--dgx-spark-x1--vllm--nvfp4--p2048-o256-c1--2026-07-07

Nemotron 3 Nano 30B-A3B on NVIDIA DGX Spark

compare

62.1tok/s

249msttft

8.2kprefill

vLLM 0.24.0NVFP4Jul 7, 2026

Run configuration

nodes

batch size

—

concurrency

context

10240

prompt / output

2042 / 256

weights nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-NVFP4 ↗

Metrics

decode tok/s

62.1

prefill tok/s

8.2k

ttft

249ms

total tok/s

58.8

inter-token

16ms

e2e latency

4.35s

req/s

0.23

peak memory

68.3 GB

avg power

29.5 W

energy/req

128.4 J

Reproduce

PATH="$HOME/venvs/vllm/bin:$PATH" VLLM_USE_DEEP_GEMM=0 TORCHINDUCTOR_COMPILE_THREADS=2 MAX_JOBS=4 vllm serve ~/models/hf/Nemotron-3-Nano-30B-A3B-NVFP4 --served-model-name Nemotron-3-Nano-30B-A3B-NVFP4 --max-model-len 10240 --gpu-memory-utilization 0.5 --max-num-seqs 8 --limit-mm-per-prompt '{"image": 0, "video": 0}' --port 8000
python3 bench/harness.py bench/configs/nemotron-3-nano-30b-a3b-nvfp4--vllm.json --rows-out data/benchmarks/nemotron-3-nano-30b-a3b.json

NVIDIA's official NVFP4 quant of Nemotron-3-Nano (nemotron_h hybrid Mamba-Transformer MoE, ~3B active). Native Blackwell FP4 on GB10.