OpenAI’s Jalapeño chip is built for fast inference at scale, benchmarks show

OpenAI's Jalapeño chip, tested on Semianalysis's InferenceX benchmark, delivered more tokens per user and greater throughput per kilowatt than current state-of-the-art hardware. The results signal progress in scaling AI inference efficiently. Jalapeño appears optimized for large-scale deployment, outperforming existing options on both performance and energy efficiency metrics, according to the benchmark data.