The DeepSeek-V4 model, released on April 24, is capable of processing vast amounts of information almost instantly. In a test, it "swallowed" 970,000 characters — roughly equivalent to seven volumes of Harry Potter and the Philosopher's Stone — and processed the query flawlessly. Such a volume is comparable to a 500-page contract, dozens of news articles, and scientific papers all at once, from which key risks and opportunities must be swiftly extracted.

But the most impressive result goes beyond mere reading. The model is capable of answering highly complex, cross-referencing questions that previously required hours of work from human analysts. For example, it was given a non-trivial question: "In which countries were new railways, built with the help of China, launched in the second half of 2025, and how did they benefit local people?"

The AI instantly analyzed thousands of publications and delivered a precise answer, highlighting projects in Latin America and Africa. Specifically, it pointed to the upgraded TAZARA railway and provided concrete figures: freight transportation time will be reduced by nearly two-thirds, and the railway's annual throughput capacity will increase to 2.4 million tons. This is the level of deep analytics now available in seconds.