Blackwell tops a new benchmark for agentic AI infrastructure

Its Blackwell platform led the first AgentPerf benchmark from Artificial Analysis.

The benchmark is aimed at agentic AI infrastructure rather than simple one-shot model calls. That matters because agents can involve long tool chains, repeated inference, and coordination across multiple steps.

As AI workloads shift from chat to autonomous execution, infrastructure comparisons are likely to focus less on peak tokens alone and more on end-to-end agent performance.