NVIDIA has achieved top performance in the latest MLPerf benchmarks for both AI model training and inference. Their new Vera Rubin platform debuted by sweeping all MLPerf Training v6.1 benchmarks and demonstrating significant generational gains in agentic AI inference throughput per watt. The platform's scalability was highlighted by efficient performance across thousands of GPUs, showcasing the integration of NVIDIA's NVLink and InfiniBand networking with optimized software. AI
IMPACT Sets new SOTA on MLPerf benchmarks, potentially accelerating AI development and deployment.
RANK_REASON NVIDIA's announcement of their new Vera Rubin platform and its leading performance in MLPerf benchmarks. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- Azure
- CUDA
- DeepSeek-V3
- GB200
- GB300 NVL72
- MLPerf
- MLPerf Endpoints
- MLPerf Training v6.1
- Nvidia
- Nvidia NVLink
- NVIDIA Quantum-X800 InfiniBand
- Spectrum-X Ethernet
- Vera Rubin
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →