SemiAnalysis· @SemiAnalysis_ · X·· 7 天前AI 评分51
AI 导读
NVIDIA 提交了首个 Rubin NVL72 跑 Kimi K3 的基准提交，在 2.8 万亿参数规模下单用户速度超过 850 tok/s，基于 vLLM 的吞吐量提升超过 3 倍。作者还提到 Rubin Software 1.0 GA 即将推出。
AI 生成摘要 · 以原文为准
正文
BREAKING NEWS🚨: NVIDIA just submitted its first Rubin NVL72 Kimi K3 submission, and it can reach over 850 tok/s/user, even at 2.8 Trillion parameters, with over 3x higher throughput on @vllm_project!
Great work by the NVIDIA team, and excited for Rubin Software 1.0 GA coming soon!
来源：SemiAnalysis · x.com