Storyline: DeepSeek's R1 model upgrade delivers major performance leap [Voice_over] DeepSeek R1, the popular AI reasoning model from the Chinese startup of the same name, is back in the headlines with the release of its first-ever upgrade. According to the company, the latest version, DeepSeek-R1-0528, has significantly improved its depth of reasoning and inference capabilities, and demonstrates outstanding performance across various benchmark evaluations, including mathematics, programming, and general logic. And for quantitative modeling, futures engineers at a securities and investment firm in Shanghai say it provides markedly faster response times. [Sound_bite] Zhang Junwu, Director, Quantitative Engineering Department, Guosen Futures: "For quantitative engineering, we conducted a comprehensive modeling test that covered the full pipeline, including code generation for client data preprocessing and drawdown calculations. We’ve observed a nearly 60 percent speed improvement, a quantum leap over previous benchmarks." [Voice_over] Other companies using the upgraded model have also confirmed the substantial improvements in its capabilities. [Sound_bite] Ge Jianing, product manager, cloud computing service provider: "The latest model shows improved programming capabilities, as demonstrated in our Minesweeper coding test. The new version, 0528, generates games with higher grid complexity while featuring more intuitive UI design that better matches user habits. Additionally, it demonstrates enhanced sustained reasoning ability, maintaining coherent output for over 20 minutes of continuous operation." [Voice_over] With a performance that rivals leading AI models like ChatGPT, Deepseek R1 has surged in popularity since its release in January, with its cost-efficiency and accessibility through open-source shocking markets around the world. [Restrictions : No access Chinese mainland]