you don't get the model when you buy the data center, & no amount of running smaller models on a tiny 200k$ "cluster" (that's like one 4 gpus node, not even 8)

you don't get the model when you buy the data center, & no amount of running smaller models on a tiny 200k$ "cluster" (that's like one 4 gpus node, not even 8) will get you remotely close to Fable 5 level performance