**[OpenAI's "Jalapeño" Takes on NVIDIA: Up to 3.6x Faster AI Inference Performance, Small-Scale Deployment by Year-End]** Beating AI News Flash: OpenAI has announced the latest test results for its self-developed AI inference chip, Jalapeño (translated as: Mexican chili pepper). According to the results, the chip outperformed NVIDIA's GB200 and GB300 superchips in the InferenceX benchmark tests. On models such as GPT-OSS 120B, DeepSeek R1, and Kimi K2.5 1T, the AI workload per watt reached 1.5 to 1.9 times that of competing products, while end-to-end latency was reduced by 1.7 to 3.6 times. OpenAI stated that Jalapeño is built on an ASIC architecture co-developed with Broadcom, specifically designed for AI inference, aiming to simultaneously improve throughput and reduce latency. OpenAI plans to deploy the chip on a small scale by the end of this year, ramp up production by 2027, and continue developing second- and third-generation chips. However, the company emphasized that it does not intend to completely replace chips from partners like NVIDIA. [Original Link]