Nvidia says its Groq 3 LPX is four times faster than Cerebras, but the math is more complicated

Nvidia has moved its Groq 3 LPX inference chip into full production, reporting speeds of 3,400 tokens per second when running Gemma 4 31B. While this performance is four times faster than the competing hardware from Cerebras, the comparison is complex because Nvidia requires a cluster of 64 accelerators to reach these speeds. In contrast, Cerebras achieves its output using only one or two units, highlighting different architectural approaches to scaling inference performance.