OpenAI said its first custom inference chip, Jalapeño, delivered faster responses and better power efficiency than competing systems in tests across several large language models...
The article requires paid subscription.
Subscribe Now