Assessing DeepSeek in hypersonic applications: Fundamental knowledge, formula invocation and automated programing

DeepSeek, as the latest large language model (LLM), has demonstrated state-of-the-art performance across diverse benchmarks. However, its potential in the hypersonic fields remains unexplored. This study aims to investigate the applicability of DeepSeek in hypersonic domain, with a focus on leveraging its chain-of-thought (CoT) reasoning capabilities. We establish a hypersonic benchmark that includes fundamental knowledge, formula invocation, and automated programing, and conduct a comprehensive comparison and evaluation of several advanced LLMs. The results show that DeepSeek-R1 performed best compared to other LLMs, achieving a 100% task completion rate on the established benchmark. It benefits from CoT reasoning mechanism, which can significantly improve hypersonic task completion rate. This work provides guidance for the design and development of LLMs specifically for hypersonic applications.