On Monday, DeepSeek unveiled an “experimental” version of V3, its foundation model first released in December. The V3.2-Exp introduces a new technique called “sparse attention mechanism” as an “intermediate step” towards the next generation of its model architecture. The method is designed to enhance efficiency while reducing training costs – a goal that aligns with China’s push to develop competitive #Artificial Intelligence products despite a lack of access to advanced Nvidia chips. The start-up said last month it was working to tailor its models for next-generation AI chips developed in #🇨🇳 ChinA.I. 🤖🧠🦾🤖 https://www.scmp.com/tech..