HOW WE DEPLOY TRILLION-PARAMETER FRONTIER MODELS ON CEREBRAS

Summary from the TLDR AI newsletter

A single Cerebras WSE-3T wafer has 44 GB of on-chip memory. Trillion-parameter frontier models need terabytes of memory for their weights alone. This post explains how the company splits a large frontier model across multiple WSE-3T wafers. It covers how models are filled into on-chip memory, how models are mapped across multiple chips, and how Cerebras handles a high number of concurrent requests.