Inside Prometheus: Meta’s Journey to a Gigawatt-Scale Supercluster
Meta's Prometheus gigawatt-scale supercluster reached phase one completion training Muse frontier models.
“an issue on a single host can slow down or stop the entire job, lowering its 'goodput'”
Meta has completed the first phase of Prometheus, a gigawatt-scale supercluster in New Albany, Ohio, targeting ~1 million GPUs. It was used to train the Muse Wonder frontier models, achieving 85%+ goodput for pre-training and 90%+ for RL runs with ~54K GPU hours MTBI. This signals Meta's serious infrastructure commitment to competing at the frontier model level, directly powering its Meta Superintelligence Labs ambitions.