[OpenAI's Self-Developed Chip Jalapeño Debuts, Achieving 1.7x the Per-Watt Throughput of GB300] According to DeepTech TechFlow, on August 26, OpenAI's self-developed inference chip Jalapeño, previously reported, has revealed its performance data for the first time. In tests with three models—GPT-OSS 120B, DeepSeek R1 670B, and Kimi K2.5 1T—Jalapeño's per-watt AI throughput exceeded that of GB200 and GB300 by 1.5 to 1.9 times. Specifically, under the DeepSeek R1 workload, its per-watt throughput reached 1.7x that of GB300. End-to-end latency was approximately 28% to 59% of the comparison systems.