DeepSeek Unveils Upgraded AI Model, Claims Performance on Par with ChatGPT and Gemini

China-based artificial intelligence company DeepSeek has announced the release of its upgraded AI model, R1-0528, boasting significantly enhanced reasoning and inference capabilities. The company claims that its latest model is narrowing the gap with top-tier competitors like OpenAI’s ChatGPT and Google’s Gemini.

According to DeepSeek, the R1-0528 model has made major strides in logical reasoning, mathematics, and programming, while also significantly reducing hallucination rates — a common issue in current AI systems. The company states that the overall performance of the upgraded model is now approaching that of leading models such as O3 and Gemini 2.5 Pro.

The debut of the original R1 chatbot in January sent shockwaves through the global AI industry, further solidifying China’s growing influence in the tech race. Despite a relatively modest training cost of $6 million, the first DeepSeek model delivered performance close to that of models trained with exponentially higher investments.

According to data from Business of Apps, DeepSeek has been downloaded 75 million times since launch and had 38 million monthly active users as of April. In a recent antitrust case, Google estimated that its Gemini model had 350 million users in March, while OpenAI’s ChatGPT reached 600 million active users in the same month.

China-U.S. AI Race Heats Up

The U.S. government is planning new restrictions on the export of advanced chip design software to China, according to a Bloomberg report. The move is aimed at limiting China’s ability to develop its domestic semiconductor manufacturing — a critical component for training and running sophisticated AI models.

Alongside DeepSeek, other Chinese tech giants like Tencent and Alibaba have also launched new AI models — T1 and Qwen3, respectively — in early 2025, accelerating the global AI race.