OpenAI’s Jalapeño chip is built for fast inference at scale, benchmarks show
TechCrunch
OpenAI's new Jalapeño chip demonstrates significant performance and efficiency advantages over current state-of-the-art inference processors in recent benchmarks.Summary
At the Hot Chips conference, OpenAI released initial benchmark results for its custom Jalapeño chip, developed in collaboration with Broadcom. Tested on Semianalysis’s InferenceX benchmark, the chip outperformed current state-of-the-art processors like Nvidia Blackwell by delivering more tokens per user and higher throughput per kilowatt. Designed to minimize data movement and bottlenecks during the prefill and communication phases, Jalapeño aims to provide fast, low-latency AI inference at scale. OpenAI plans an initial rollout in late 2026, with broader deployment expected in 2027 as part of a multigenerational hardware strategy.
(Source：TechCrunch)