Actually, the 5 million figure is for the compute cost for the base 600B parameter model. Training R1 was just 8000 steps of reinforcement learning, so I expect

Actually, the 5 million figure is for the compute cost for the base 600B parameter model. Training R1 was just 8000 steps of reinforcement learning, so I expect that the vast, vast majority of the training cost is already included in the pretraining stage.