Notably, MXFP4 was introduced at the (much less costly) supervised fine-tuning stage after pretraining, so the number of B200/B300 GPUs could be relatively small in comparison to the number of H200 GPUs used during pretraining (or maybe not, who knows).
Comments
Notably, MXFP4 was introduced at the (much less costly) supervised fine-tuning stage after pretraining, so the number of B200/B300 GPUs could be relatively small in comparison to the number of H200 GPUs used during pretraining (or maybe not, who knows).
(Kimi K3 tech report section 4.1.1 https://arxiv.org/pdf/2607.24653)