DeepSeek Upgrades AI Model V3, Boosts Performance Globally

Chinese AI startup DeepSeek has launched an upgraded version of its large language model, DeepSeek-V3-0324, which significantly enhances reasoning and coding capabilities and is now available under the MIT open-source license. This model, which boasts 685 billion parameters, outperformed its predecessor and other models like OpenAI's in various benchmarks, highlighting its competitive edge in the AI landscape. The upgrade follows the company's earlier successful R1 model, which gained attention for its efficiency and performance on limited budgets. DeepSeek's rapid growth is evident, with the app becoming the top downloaded in over 156 countries and averaging 22.15 million active users daily. Despite challenges such as server capacity issues and concerns over ties to the Chinese government, DeepSeek aims to increase its global presence and launch additional AI systems. The company’s advancements underscore China's escalating role in the global AI race, prompting competitors to accelerate their innovations.