DeepSeek, a China-based synthetic intelligence firm, has introduced an improve to its AI chatbot, saying it could now supply enhanced general logic, arithmetic and programming with a decreased hallucination price.
Based on DeepSeek, the upgraded mannequin — DeepSeek-R1-0528 — has “considerably improved its depth of reasoning and inference capabilities.” The startup mentioned the mannequin’s general efficiency is now “approaching that of main fashions, equivalent to O3 and Gemini 2.5 Professional.”
DeepSeek’s debut of its R1 chatbot in January despatched shockwaves by way of the AI trade and additional established China as an AI pressure. The corporate’s first AI mannequin had a coaching price of $6 million and comparable efficiency to main AI fashions skilled on considerably bigger sums of capital.
Based on information from Enterprise of Apps, DeepSeek has been downloaded 75 million occasions since its launch and had 38 million month-to-month energetic customers (MAU) as of April. In a current antitrust lawsuit, Google estimated that Gemini reached 350 million energetic customers in March, whereas OpenAI’s ChatGPT claimed 600 million energetic customers in the identical month.
Associated: China’s DeepSeek launches new open-source AI after R1 took on OpenAI
Chinese language-American AI race heats up
America authorities is planning to limit the sale of superior chip design software program to China. Based on a Bloomberg report, the transfer seeks to restrict China’s means to advance its home semiconductor manufacturing capabilities.
Semiconductors are vital for a variety of applied sciences, together with AI, the place they function the {hardware} spine for coaching and operating advanced fashions.
New China AI fashions, equivalent to Tencent’s T1 and Alibaba’s Qwen3, have additionally emerged within the first few months of 2025, spurring the AI race alongside.
Journal: AI Eye: 9 curious issues about DeepSeek R1