NVIDIA Sweeps MLPerf Training v6.1 as Vera Rubin Debuts in Training and Endpoints
- Published
- Source
- NvidiaNews
NVIDIA says its Vera Rubin platform delivered leading performance across training, post-training and agentic inference in its MLPerf debut. The company swept every MLPerf Training v6.1 benchmark and made its first MLPerf Endpoints submission, reporting more than 10x generational gains in throughput per watt for agentic AI. The source describes throughput per watt as a measure of how much agentic work a system delivers for each watt of power.
The Training results covered large language models, image generation, recommender systems and reinforcement-learning post-training; NVIDIA was the only platform with results on every benchmark in this round. Vera Rubin NVL72 delivered up to 2x more performance per GPU than NVIDIA GB300 NVL72, according to the article. On DeepSeek-V3, Vera Rubin NVL72 scaled from 256 to 1,024 GPUs at about 97% efficiency, reducing training time from 16 minutes to about four. NVIDIA also ran the benchmark on 10,240 GB300 GPUs, while Microsoft Azure matched that scale using 10,240 NVIDIA GB200 GPUs.
NVIDIA’s reported results make efficient scaling and power use key points for enterprise evaluation: adding GPUs accelerates training only when they work together effectively, while agentic workloads generate tokens across repeated reasoning and tool-use steps. The article attributes scaling to NVLink within systems and Quantum-X800 InfiniBand and Spectrum-X Ethernet between systems, alongside optimized software. It says Vera Rubin performance is expected to keep climbing through continuous optimization of NVIDIA’s CUDA-powered software stack; no further timeline was disclosed.
Original report
NVIDIA Sweeps MLPerf Training and Endpoints as Vera Rubin Debuts With Leading Performance
NvidiaNews · NVIDIAThis briefing is an independently written summary based on publicly available reporting and is provided for industry information and news discovery. The original report and source publication are credited and linked where applicable. RobotToday does not claim ownership of third-party source material.
Rights concerns? Contact [email protected] with the relevant URL and details. We will review the matter and take appropriate action where warranted.
More related news
- 1
- 2
- 3
- 4
- 5
- 6
-
7
RobotToday Weekly November 9–13, 2026 RobotToday · 3 days ago
-
8
RobotToday Weekly November 2 – 6, 2026 RobotToday · 1 week ago
- 9
- 10
- 11
- 12
Related Suppliers
NVIDIA Robotics
NVIDIA robotics division delivering Isaac simulation and training tools, Jetson edge compute modules, and Omniverse digital twin frameworks for physical AI.
Deci AI (Acquired by NVIDIA)
Israeli deep learning optimization startup acquired by NVIDIA in May 2024 for ~$300M; known for AutoNAC neural architecture search technology.
Nscale
London-based Nscale runs a full-stack AI/GPU cloud platform and on Aug 1, 2026 agreed to acquire Anyscale for $1.65 billion.
Beijing TT Aviation Technology Co., Ltd.
Beijing UAV maker founded 2008 with multirotor, VTOL and EDU drones for agriculture, security and survey, and 36+ pilot training centres.
MoviTHERM - Advanced Thermography Solutions
Irvine, California thermal-imaging integrator founded 1999; FLIR Premium Partner building fire-detection, NDT, and condition-monitoring camera systems.
SISCO
Distributor of test and measurement equipment founded 2006, offering lab instruments, environmental meters, gas detectors, and industrial tools globally.
Fireworks AI
Redwood City AI infra startup (2022) serving open-source LLMs like Kimi K3 via API, used by Cursor, Vercel and Notion.
Alpha Robotics 苏州阿尔法机器人有限公司
Suzhou, China commercial service robot maker founded 2006 with a Tokyo office; ships delivery, AMR and disinfection robots to 8,000 customers in 75 countries.
Modular
Modular built the Mojo language and MAX AI-inference platform for GPUs/CPUs before Qualcomm completed its acquisition of the firm in July 2026.
Cerebras Systems
Sunnyvale, California AI-compute company founded 2015; builds wafer-scale WSE-3/CS-3 chips and, since August 2026, the CS-4 accelerator.
nLink
Norwegian developer of mobile robots for dirty and dangerous tasks, based in Sogndal, that collaborates with Hilti AG on construction robots.
Quantum Devices, Inc.
US manufacturer of high-performance optical encoders and silicon photodiodes for motion feedback, plus LED systems for photodynamic therapy.
Robotics needs a service framework.
RSF defines a common language for robot service capability, lifecycle operations, certification pathways, and service-provider networks.
Join the RobotToday community on LinkedIn
Daily robotics news, in-depth analysis, conference highlights, and discussions with professionals worldwide.