OpenAI reports Jalapeno chip benchmarks against Nvidia GB300

OpenAI said its Broadcom-developed Jalapeno inference chip outperformed Nvidia's GB300 on AI work per unit of power and response speed. The company said the chip operates at 700 watts and is intended to combine high-throughput and low-latency inference work. Its tests used one small OpenAI open model and third-party models from DeepSeek and Moonshot. OpenAI said Jalapeno was not tested against Nvidia's newer Vera Rubin hardware and cannot train models. The figures are company-reported benchmarks rather than independent results.