China’s DeepSeek Unveils Upgraded AI Model, Challenging US Tech Giants
Chinese AI startup DeepSeek has announced a significant upgrade to its R1 model, named DeepSeek-R1-0528, claiming enhanced capabilities in mathematics, programming, and general logic while reducing AI ‘hallucinations.’ The announcement positions the new model as a formidable competitor to industry leaders like OpenAI’s o3 and Google’s Gemini 2.5 Pro.
DeepSeek stunned the global tech community in January with the original R1, a model that rivaled the performance of systems developed by much larger U.S. companies, reportedly at a fraction of the cost. Built for under $6 million using less-advanced Nvidia H800 chips, R1’s efficiency challenged assumptions about the necessity of massive computational investments, triggering a $1 trillion stock market drop, including Nvidia’s $589 billion loss.
The upgraded DeepSeek-R1-0528 builds on this foundation, offering what the company claims is ‘greater depth of reasoning’ and performance approaching that of the world’s top AI models. ‘Its overall performance is now approaching that of leading models,’ DeepSeek stated, signaling its ambition to compete at the forefront of the global AI race.
The timing of the DeepSeek-R1-0528 announcement was notable, coming just hours before Nvidia Corp., the Santa Clara-based leader in AI chips, released its latest financial report. Nvidia’s shares took a hit following R1’s initial launch due to concerns about reduced AI infrastructure costs, but have since recovered as global investment in AI data centers remains robust.
DeepSeek’s rapid progress highlights China’s accelerating role in the global AI arms race. Industry analysts note that the startup’s ability to deliver high-performing models at lower costs could reshape the economics of AI development, challenging the dominance of U.S. tech giants.
