Search

OpenAI and Broadcom unveil Jalapeño, a chip built for LLM inference

The accelerator targets the work of producing model responses, with initial deployment planned for the end of 2026.