NVIDIA Groq 3 LPX: Record-Breaking Speed for AI Agents

NVIDIA Groq 3 LPX is now in full production, offering world-class speed for agentic coding and other latency-sensitive AI tasks. This interactive AI inference accelerator extends the NVIDIA Vera Rubin NVL72 systems, dramatically increasing token generation rates. What's interesting is that Groq 3 LPX delivered a record 3,400 output tokens per second in benchmarking, running an open-source agentic model. This performance is the fastest ever recorded for that model. It provides four times faster responsiveness for agents and latency-sensitive workloads compared to its nearest alternative. Nebius is the first AI cloud to adopt this new technology. The bottom line: this new accelerator helps AI agents complete complex tasks and coding much faster, potentially transforming how intelligence is produced.