Cognition Becomes First Customer for NVIDIA Vera Rubin, Boasting 4.8x Token Throughput

Core summary

One sentence to understand this update

Cognition is the inaugural customer to deploy NVIDIA Vera Rubin, powered by CoreWeave, with the new chips reportedly delivering 4.8 times more token throughput than GB200 for SWE-2 inference.

Impact & opportunity

What this could mean

This signals a significant leap in AI inference capabilities, offering builders the potential for massive performance gains in computationally intensive tasks like code generation and complex reasoning.