精选理由
Cursor 说 MoK 已在数万 GPU 上跑生产，比旧 DeepEP 栈端到端快 1.41 倍，搞大规模训练的可以看。
Cursor 宣布其 MoK 方案已投入训练任务，覆盖数万张 GPU。相比此前基于 DeepEP 的技术栈，端到端训练吞吐提升至 1.41 倍。MoK 已在生产环境验证，用于 Cursor 的大规模训练基础设施。
原文 · Cursor
MoK now powers training across tens of thousands of GPUs at Cursor. In production, it raised end-to...
MoK now powers training across tens of thousands of GPUs at Cursor. In production, it raised end-to-end training throughput by 1.41x over our previous DeepEP-based stack. 💬 2 🔄 1 ❤️ 86 👀 6598 📊 9 ⚡