Mixture of experts

A single-file C engine called Colibrì streams GLM-5.2's 744B mixture-of-experts weights off an NVMe drive to run the full model on 25 GB of RAM at 0.05-2 tokens/second. Here's how it works, what Hacker News made of it, and how to check on a queued run from your phone with Pinggy.

Understand Mixture of Experts (MoE) architecture in LLM models - how it works, why it's efficient, and which models use it. Learn about sparse activation, routing, and performance benefits.