**[OpenAI Claims First Self-Developed Inference Chip Jalapeño Surpasses NVIDIA GB300: AI Output Per Watt Leads by 1.5 to 1.9 Times, Latency Reduced by Up to 3.6 Times]** Beating AI Newsflash: OpenAI has released the first batch of test data for its custom inference chip, Jalapeño, claiming performance surpasses NVIDIA GB300. The chip achieves Pareto optimality on three publicly available external models: GPT-OSS 120B, DeepSeek R1 670B, and Kimi K2.5 1T. Compared to the best existing commercial systems, peak AI output per watt is 1.5 to 1.9 times higher, while end-to-end latency is reduced by 1.7 to 3.6 times; in high-interactivity intelligent agent scenarios, the advantage expands to 2.1 to 4.1 times. Jalapeño has a rated power of 700 watts, with measured sustained power not exceeding 550 watts. OpenAI emphasizes that under intelligent agent workloads, the true cost should be measured by "AI workload completed per unit of power consumption" rather than single-chip performance. Jalapeño was designed and taped out in just 9 months, with AI deeply involved in circuit optimization and validation. Leveraging Codex and GPT-Astra, the team optimized three open-weight models not included in the original plan to high performance within two months, with AI-generated implementations for certain modules being 1.5 to 1.8 times faster than human-written implementations. OpenAI plans to deploy Jalapeño to its proprietary compute infrastructure by the end of the year while continuing large-scale use of external accelerators like NVIDIA. This marks the first generation in a multi-generation chip roadmap, with Gen 2 already in deep development and Gen 3 taking shape. [Original Link]