‘We made a chip and it is fast’: OpenAI’s Jalapeno shows big gains in speed
OpenAI says its first custom inference chip can deliver more AI work per watt while reducing response times, as it moves towards deploying its own silicon alongside Nvidia and other accelerators
