Case Study: Prima Mente boosts pre-training throughput with Nebius reserved GPU clusters

Prima Mente boosts pre-training throughput 12% with Nebius

Prima Mente, an AI biology company, faced the challenge of scaling its biological foundation model, Pleiades, to process massive datasets for neurodegenerative disease research. Their existing infrastructure created friction and overhead, hindering the speed of their critical AI training cycles and experiments. To overcome this, they sought a more stable and reliable cloud partner in Nebius.

By implementing a dedicated 32-GPU reserved cluster from Nebius, Prima Mente achieved a 12% increase in pre-training throughput, processing 1.19 million tokens per second across 16 nodes. This performance boost from Nebius was a critical step that enabled the company to efficiently scale its model from 7B parameters toward its goal of exceeding 100B parameters, accelerating their path to new diagnostics and therapies.

Prima Mente