Colibri Runs Massive MoE Models on CPU+SSD, Challenges GPU Dominance

Colibri, a new project, runs massive Mixture of Experts models on ordinary CPUs and SSDs by loading only needed experts into memory. It treats storage as an active inference participant, not just a weight repository. The approach could democratize large model access by reducing hardware requirements.