AI 摘要
在Cerebras上以最大上下文窗口运行单个深度编码模型,仅支持256个并发用户就需要24套系统(2400万美元资本支出)。在这个规模下,1亿美元在标准GB300机架中能获得高得多的内存带宽。
Running a single deep coding model at max context on Cerebras requires 24 systems ($24M Capex) just to support 256 concurrent users. At that scale, $100M gets you way more memory bandwidth in standard GB300 racks.