Back previous/next report: k j

Google GB300-NVL72_GB300-288GB_aarch64x72_TRTLLM

Submitted by Google LLCon 2026-10-16 Published on 2026-11-04

Result statement

Google LLC ran GPT-OSS 120B on GB300-NVL72_GB300-288GB_aarch64x72_TRTLLM and achieved a verified 948,368.9 tokens/second System TPS at 509.0ms TTFT P90 in the Standardized division of the Available category on November 4, 2026 .

[1] MLPerf Endpoints v1.0 GPT-OSS 120B, published November 4, 2026; retrieved from results.mlcommons.org on November 18, 2026. Result —, run 8: concurrency = 22,528 queries, TTFT P90 = 509.0ms, System TPS = 948,368.9 tokens/sec, interactivity = 42.1 tokens/sec per user and total provisioned system power = 151.82 kW (MLC Estimated Power). Result verified by MLCommons Association. The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information.

Submission Info
Google LLC
d6b88dff-6bbb-4107-8d2e-1090decd4d53
Submission Status PUBLISHED
2026-10-16
2026-11-04
Version v1.0
Division Standardized
Deployment Client on Premises
Model
Model Name GPT-OSS 120B
Division Standardized
Model Precision —
Model Link —
Calibration Link —
Model Notes —
Datasets
Dataset Name perf
Dataset Type performance
Average Input Tokens —
Dataset Link —

System TPS vs. Tokens/s per User

0200k400k600k800k1.0MSystem TPS100200300400Tokens/s per User

System TPS vs. Concurrency

0200k400k600k800k1.0MSystem TPS05,00010,00015,00020,000# of queries

Time to First Token P90 vs. Concurrency

200400600800TTFT P90 (ms)05,00010,00015,00020,000# of queries

Tokens/s per User vs. Concurrency

100200300400Tokens/s per User05,00010,00015,00020,000# of queries
SUT Summary
NVIDIA
GB300-NVL72_GB300-288GB_aarch64x72_TRTLLM
Available
System Category Datacenter · —
18 nodes x 4 NVIDIA GB300 (72 GPUs, one NVL72 domain)
1
18
Max Supported Concurrency 22,528
Node Type ID 0
Node Count 18
Software
Framework TensorRT-LLM 1.3.0rc14 (disaggregated prefill/decode)
TensorRT-LLM 1.3.0rc14 (disaggregated prefill/decode)
TensorRT-LLM
NVIDIA 580.173.02 (GKE 'default' COS driver; CUDA 13.0 driver API)
Link to Container nvcr.io/nvidia/tensorrt-llm/release:1.3.0rc14
Operating System Container-Optimized OS, kernel 6.12.94+, GKE 1.35.8-gke.1036000
Filesystem ext4 on node-local NVMe SSD (/mnt/stateful_partition/kube-ephemeral-ssd)
Other Software MLCommons endpoints client ghcr.io/mlcommons/endpoints:e9f427c; LiveCodeBench scorer ghcr.io/mlcommons/endpoints:9d749906-livecodebench; containerd 2.2.7
Software Notes —
Hardware

Processor

NVIDIA Grace (Arm Neoverse V2)
0
—
VCPUs Per Processor 144

Accelerator

—
0
—
—
—
—

Host / Storage

~943 GiB per node (kubectl capacity 988634624Ki)
Memory Configuration 960 GB LPDDR5X (2x NVIDIA Grace CPU integrated memory)
Storage Capacity ~12.5 TB local SSD per node (kubectl ephemeral-storage)
Storage Type —

Networking

4x NVIDIA ConnectX-7 + 1x Google Cloud Titanium NIC (a4x-maxgpu-4g-metal)
Google Cloud Titanium / RoCE + NVLink 5 (NVL72)
Network Topology —

Other

Other Hardware —
Cooling Liquid-cooled
Hardware Notes —
Run Data
Field NameRun 1Run 2Run 3Run 4Run 5Run 6Run 7Run 8
Run Date10/16/202610/16/202610/16/202610/16/202610/16/202610/16/202610/16/202610/16/2026
Concurrency1735701402721,0243,84022,528
System Tokens/Second6,700.912,762.323,201.741,220.671,891.3192,483.8445,733.1948,368.9
Tokens/Second per User394.2364.6331.5294.4264.3188.0116.142.1
TTFT P90 (ms)89.787.7102.2147.6610.9705.9849.8509.0
Utilization0.7%1.3%2.4%4.3%7.6%20.3%47.0%100.0%
Configuration Summarygpt-oss-120b-c17 · submissiongpt-oss-120b-c35 · submissiongpt-oss-120b-c70 · submissiongpt-oss-120b-c140 · submissiongpt-oss-120b-c272 · submissiongpt-oss-120b-c1024 · submissiongpt-oss-120b-c3840 · submissiongpt-oss-120b-c22528 · submission
Config Typesubmissionsubmissionsubmissionsubmissionsubmissionsubmissionsubmissionsubmission
Config Region————————
Target QPS————————
Result Statecompletecompletecompletecompletecompletecompletecompletecomplete
Measured Accuracy Score————————
Benchmark Versionunknownunknownunknownunknownunknownunknownunknownunknown
Result Version0.1.00.1.00.1.00.1.00.1.00.1.00.1.00.1.0
Archive URIgs://mlperf-endpoints-blob/runs/6882566f-7eb0-4cdc-b183-bd1487b7c0c3/archive.tar.gzgs://mlperf-endpoints-blob/runs/3c9a8a07-5451-42da-adee-1876397c2e64/archive.tar.gzgs://mlperf-endpoints-blob/runs/d130215f-124d-4e5d-8195-262c3d7d210b/archive.tar.gzgs://mlperf-endpoints-blob/runs/3e6761dc-e335-4c5d-998f-515f64f59a7b/archive.tar.gzgs://mlperf-endpoints-blob/runs/712660c4-9314-4633-83a6-f62650a5ffc8/archive.tar.gzgs://mlperf-endpoints-blob/runs/30e2e482-54b0-4053-a93e-7c5740c6ed9f/archive.tar.gzgs://mlperf-endpoints-blob/runs/0d350459-3835-49c9-9ba8-d4462719cdd9/archive.tar.gzgs://mlperf-endpoints-blob/runs/caacd918-5f8a-4cec-8557-107cc85429f5/archive.tar.gz
QPS5.459510.373518.852633.437858.3776156.4375362.2483770.8032
Total Output Tokens7,850,34147,213,21539,357,30339,423,52694,519,024251,832,572566,642,8921,888,658,721
Average Output Tokens1,227.381,230.281,230.681,232.761,231.491,230.421,230.461,230.36
Run Duration (s)1,171.543,699.421,696.31956.401,314.751,308.331,271.261,991.48
Total Queries6,39638,37631,98031,98076,752204,672460,5121,535,040
Total Queries Issued6,39638,37631,98031,98076,752204,672460,5121,535,040
Total Queries Failed00000000
Time to First Token (TTFT) (ms)
Minimum32.229.629.830.229.933.541.859.2
Average82.868.879.2105.6243.2384.9580.2516.0
P5061.860.963.479.393.0322.3500.9325.7
P9089.787.7102.2147.6610.9705.9849.8509.0
P95104.9101.1118.1175.7953.7867.0979.5587.1
P991033.9130.5155.5248.21812.31196.91474.37663.4
P9993037.41277.93071.23658.93581.83935.98868.021163.8
Maximum3466.63575.23687.94006.44270.45340.310737.924821.0
Time Per Output Token (TPOT) (ms)
Minimum0.811.411.141.571.641.874.7511.2
Average2.462.692.953.303.584.998.0923.2
P502.462.672.943.293.565.008.1523.0
P902.532.773.093.433.725.068.4524.3
P952.562.833.143.473.755.088.5024.6
P992.612.933.243.543.775.118.5625.0
P9992.663.073.383.633.825.178.6325.2
Maximum2.783.193.533.703.925.8116.543.5
Query Latency (ms)
Minimum1427.71318.11461.11673.81663.62266.43609.01755.4
Average3108.83371.73706.94175.04646.96521.010529.928978.2
P503059.43333.23641.74108.94563.46447.110387.228565.6
P903913.44227.74708.75259.85882.88105.713127.936334.5
P954194.04522.35061.05651.36328.98668.514059.338981.4
P994877.65158.85814.36484.17321.39921.416253.345228.7
P9995906.96166.47067.68000.48660.911885.820242.554481.6
Maximum7108.97947.09114.712460.311947.916620.828441.783625.5