OpenAI's Jalapeño chip claims 3.6x faster AI inference than Nvidia
OpenAI has published new benchmark results for its in-house AI inference chip, Jalapeño, and the company says the chip beats Nvidia's GB200 and GB300 superchips on the InferenceX suite. According to OpenAI, Jalapeño delivers 1.5 to 1.9 times more AI work per watt on models like GPT-OSS 120B, DeepSeek R1, and Kimi K2.5 1T, while cutting end-to-end latency by 1.7 to 3.6 times.
Jalapeño uses an ASIC architecture developed with Broadcom, built specifically for AI inference to boost throughput and reduce latency. OpenAI plans a small-scale deployment by the end of this year, expanding production in 2027, and is already working on second- and third-generation chips. Still, the company stresses that Jalapeño won't fully replace chips from partners like Nvidia.