DeepSeek, a Chinese AI startup, has established itself as a formidable challenger to Western AI dominance with its latest R1-0528 model update, which now performs at levels approaching industry leaders OpenAI and Google.
The upgraded model, while described by DeepSeek as a "minor version upgrade," delivers substantial improvements across key capabilities. In mathematical reasoning, the model's accuracy on the challenging AIME 2025 test jumped from 70% to 87.5%, demonstrating enhanced problem-solving abilities. This advancement stems from the model's improved reasoning depth, now utilizing approximately 23,000 tokens per query compared to 12,000 in the previous version.
Beyond mathematics, R1-0528 shows significant gains in programming proficiency. On the Codeforces programming challenge, the model achieved a rating of approximately 1930, up from 1530 previously—a 400-point improvement reflecting superior code generation and problem-solving capabilities. Its performance on comprehensive coding tests has also improved substantially.
The model's progress aligns with findings from Stanford University's 2025 AI Index Report, which documents a dramatic narrowing of the performance gap between top US and Chinese AI models. According to the report, the difference between leading American and Chinese models on major benchmarks shrank from double digits in 2023 to near parity by early 2025. On the Chatbot Arena Leaderboard, the gap between top US and Chinese models decreased from 9.26% in January 2024 to just 1.70% by February 2025.
Industry leaders have taken notice of this shift. Nvidia CEO Jensen Huang acknowledged in May that "DeepSeek and [Alibaba's] Qwen from China are among the best open-source AI models released freely. They've gained traction across the U.S., Europe and beyond."
DeepSeek's achievement is particularly notable as it comes from a company that began as a relatively small startup. The model is available under a permissive license, with both the full version and a smaller "distilled" variant that can run on less powerful hardware, making advanced AI capabilities more accessible to developers worldwide.