DeepSeek 知乎披露 V4.1 Flash 服务数据：单集群 155 万亿 token/天

swyx @swyx原推发布于 2026.10.21

在 X 上打开原推

deepseek posted way more alpha on their zhihu page, if you only read twitter you are reading the laowai approved stuff "Based on this framework, we have implemented a high-throughput, low-latency large-scale inference service for DeepSeek V4.1 Flash. Currently, the service throughput of a single cluster is 155 trillion tokens/day, and the overall revenue is $38 per card per hour (based on off-peak prices)." it costs roughtly $5-10 per gpu hour so depseek is claiming 80% serving margin at THEIR prices lololololol

这条是引用着另一条写的：「DeepSeek 开源推理框架 DeepInfer，支撑 V4.1 Flash 生产服务」