LaunchDevelopersNovember 18, 2026

Cerebras launches disaggregated inference with GPUs and wafer-scale engines

Read original source →cerebras.ai

GPUs now handle prefill and Cerebras wafer-scale engines handle decode, delivering 5x more tokens per megawatt at the same Cerebras speeds. The setup is in production and available to infrastructure partners for their own datacenters.

2 sources

More stories today

Open the live feed