DeepSeek unveils the R1-0528 model with better reasoning, claiming it’s nearing the performance of OpenAI and Google’s top AI systems.
DeepSeek, an artificial intelligence company headquartered in China, has recently disclosed an upgrade to its AI chatbot.
The company claims that the chatbot can now provide a reduced rate of hallucinations and an improved level of overall logic, mathematics, and programming.
The upgraded model, DeepSeek-R1-0528, has “substantially enhanced its inference and reasoning capabilities,” according to DeepSeek.
The startup stated that the model’s overall functionality is now “approaching that of leading models, such as O3 and Gemini 2.5 Pro.”
In January, DeepSeek’s R1 chatbot was introduced, causing a significant impact on the AI industry and further solidifying China’s status as a leader in the field.
The company’s initial AI model had a training cost of $6 million and exhibited comparable performance to leading AI models trained on substantially larger sums of capital.
As of April, DeepSeek had 38 million monthly active users (MAU) and had been downloaded 75 million times since its debut, as per data from Business of Apps.
Google estimated that Gemini had 350 million active users in March in recent antitrust litigation, while OpenAI’s ChatGPT claimed 600 million active users in the same month.
AI Competition Between China, United States Is Intensifying
The transfer of sophisticated chip design software to China is anticipated to be restricted by the United States government.
According to a Bloomberg report, this action is intended to limit China’s capacity to enhance its domestic semiconductor manufacturing capabilities.
Semiconductors are indispensable for an extensive array of technologies, including artificial intelligence (AI), where they function as the hardware foundation for the development and execution of intricate models.
The AI race has been further fueled by the emergence of new China AI models, including Tencent’s T1 and Alibaba’s Qwen3, in the first few months of 2025.
