Together AI 推出 Blackwell GPU 集群，训练速度提升90%
原标题：Kernel Collection
AI 摘要
Together AI 宣布推出基于 NVIDIA Blackwell 平台的 GPU 集群，并配套优化的 AI 加速软件栈 Together Kernel Collection。该集群在训练 70B 参数模型时速度比 H100 快 90%，达到每节点每秒 15,200 tokens。Together AI 利用 ThunderKittens 框架开发了自定义 FP8 内核，其注意力内核比 FlashAttention-3 快 1.8 倍。此外，公司计划部署数万台 HGX B200 服务器和 GB200 NVL72 机架级解决方案，并推出独家试用计划，邀请八支 AI 团队参与。
正文节选
Today we are announcing immediate access to Together GPU Clusters accelerated by the NVIDIA Blackwell platform, and an accompanying AI acceleration stack optimized for the latest GPU architecture. Together GPU Clusters featuring NVIDIA HGX B200 are turbocharged with Together Kernel Collection to deliver unprecedented performance: 90% faster training than NVIDIA HGX H100, achieving 15,200 tokens/second/node on a training run for a 70B parameter model. Our research team has achieved these incredib