How OpenAI Used Its Own LLMs to Design Its Jalapeño Chip
IEEE Spectrum
Read full postOpenAI unveiled its first AI accelerator chip, Jalapeño, delivering 13.4 petaflops with advanced memory and reduced latency compared to Nvidia's GB300. The chip was designed rapidly in under 20 months using OpenAI's large language models to accelerate the process, with a small team and Broadcom partnership handling physical design.
