NVIDIA · NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)
DeepSeek-R1 (MLPerf reference) · deepseek-r1 · Server · run 01a00b4a-2e6c…
valid · Reported evidence
Measurements
output_token_throughput_tps · reported400,437 tokens_s
DeepSeek-R1 (MLPerf reference) · deepseek-r1 · Server · server · 4× NVIDIA GB300 × 18 nodes · mlperf_closed_99 · observed 2026-03-26
Identity
modelDeepSeek-R1 (MLPerf reference)
stackTensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM
workloaddeepseek-r1 · Server
topology4× NVIDIA GB300 × 18 nodes
harnessmlperf_loadgen v6.0
quality policymlperf_closed_99
comparison keysha256:f4524df374c23138…
sha256:078f27c9cf4c1557475ba45…
Launch configuration
dtypefp4
notesNVIDIA GB300 NVL72, 1400W TGP Per GPU
systemNVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)
softwareTensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0
availabilityavailable
Workload
namedeepseek-r1 · Server
streamingyes
load generationopen_loop
Canonical manifest
Show manifestHide manifest
{
"kind": "ServingRun",
"spec": {
"status": "valid",
"harness": {
"name": "mlperf_loadgen",
"version": "v6.0"
},
"topology": {
"nodeCount": 18,
"acceleratorModel": "NVIDIA GB300",
"acceleratorVendor": "nvidia",
"acceleratorsPerNode": 4
},
"observedAt": "2026-03-26T19:39:54Z",
"modelDigest": "sha256:9c4d3f416e476a367a8351b29d53c726179e98b9f039335873b937afb40e7c29",
"measurements": [
{
"unit": "tokens_s",
"value": 400437,
"metric": "output_token_throughput_tps",
"statistic": "reported"
}
],
"sourceNative": {
"source": "mlperf-inference",
"metrics": {
"performance_units": "Tokens/s",
"performance_result": 400437
},
"category": "datacenter",
"division": "closed",
"scenario": "server",
"benchmark": "deepseek-r1",
"externalId": "v6.0/6.0-0078"
},
"qualityPolicy": "mlperf_closed_99",
"workloadDigest": "sha256:51ac3e54124be56b08b48d2e88fb8878fda1ff10c64431a52e66b4530fc4d7d9",
"configurationDigest": "sha256:04ff27d0c4a77d2ee2422a71ca2ac671de19d9288230a2a85154a09f1045973f"
},
"metadata": {
"name": "mlperf-v6-0-6-0-0078",
"title": "NVIDIA · deepseek-r1 · Server",
"sourceRefs": [
{
"url": "https://github.com/mlcommons/submissions_inference_v6.0/tree/main/closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT"
}
]
},
"apiVersion": "kernelindex.dev/v1alpha1"
}Cite this result (permalink, digest, access date)
Report an issue with this serving run
MLPerf Inference · closed/datacenter · entry v6.0/6.0-0078 · shown unmodified as publishedAll serving results →JSON