DeepSeek Made Its Model 1.9x Bigger and 4x Cheaper to Remember.
中文摘要
DeepSeek 在将模型容量扩大 1.9 倍的同时，将记忆成本降低了 4 倍。
English Summary
DeepSeek increased its model capacity by 1.9x while reducing memory costs by 4x.
Original Excerpt
Here’s the Arithmetic Behind. Continue reading on Medium »