MLPerf Training v6.0: Lambda delivers fastest LLM training on NVIDIA GB300 NVL72 and fastest MoE training on NVIDIA HGX B200

What this post added

This post reports Lambda's specific results for MLPerf Training v6.0, detailing the fastest LLM training times on NVIDIA GB300 NVL72 and the fastest MoE training times on NVIDIA HGX B200. It quantifies the 18.7% improvement in training speed for Llama 3.1 8B on GB300 NVL72 due to software optimizations and provides convergence times for GPT-OSS-20B on both GB300 NVL72 and HGX B200 systems. It also lists system specifications for the tested configurations and compares Lambda's performance against other submissions.