AMD releases Instella-MoE-16B-A3B, open MoE LLM with 2.8B active params

AMD's Instella-MoE-16B-A3B is a fully open Mixture-of-Experts LLM with 16B total parameters and 2.8B active per token. It was trained from scratch on Instinct MI300X and MI325X GPUs, with AMD releasing weights from every training stage.