This is the largest model ever trained through decentralized compute/cooperation It needs to win. If LLM training isn't decentralized soon we will have a very big problem