Hardware-based AI inference engine that delivers responses at 500+ tokens per second—the fastest in the industry. Groq