Pinned
Since announcing Jalapeño, our first custom inference chip, we’ve been testing it and the system around it.
The results show a major advance: more intelligence from every watt and faster responses, delivering both higher throughput and lower latency in one architecture without

