Firehose

OpenAI's LLMs accelerated the design of its Jalapeño chip from first architecture concept to first silicon in under 20 months, with a small design team averaging fewer than 100 people. The team used Accelerated Hardware Synthesis (XLS) and OpenAI's internal LLMs, including models like o3 and precursors to GPT-6 Astra, to design the chip, which delivers up to 13.4 petaflops of 4-bit compute and accesses 232 gigabytes of memory. The team's workflow has a lot of places where AI is being introduced for the second-generation chip. AI summary