As Investing.com notes, tests have shown significant gains in both speed and energy efficiency. Jalapeño breaks the traditional trade-off between throughput and latency, delivering substantial improvements over leading commercial AI hardware solutions, the publication explains.
The company tested Jalapeño using InferenceX—a public benchmark from SemiAnalysis—comparing it to leading commercially available AI systems. The tests covered three models: GPT-OSS 120B, DeepSeek R1 670B, and Kimi K2.5 1T.
The test results showed that Jalapeño delivered 1.5–1.9 times more AI computations per watt at peak throughput and 1.7 to 3.6 times lower end-to-end latency compared to competing systems across all three models. For highly interactive workloads, performance was 2.1 to 4.1 times higher.
OpenAI emphasized that the chip’s performance advantage became even more pronounced on the company’s internal state-of-the-art models, indicating that the architecture becomes more efficient as workloads increase.
It is important to note that the development process, from initial design to wafer production, took nine months. OpenAI used its own AI models to design and optimize the chip: earlier generations assisted during the design phase, while newer models accelerated optimization and programming.
Follow our updates
