China's DeepSeek Lab Unveils Next-Generation AI Models

China's DeepSeek laboratory has officially introduced its latest large language models, DeepSeek V4 Flash and V4 Pro. According to Zamin.uz, these new developments are the logical successors to the V3.2 and the popular R1 models released last year.

The next-generation models feature significant structural improvements, enabling them to analyze up to one million characters of text simultaneously. This capability allows users to work efficiently with massive documents, books, and complex programming codes.

The V4 Pro model boasts a total of 1.6 trillion parameters, making it the world's largest open-source AI system to date. For comparison, it records significantly higher results than its close competitors, such as Moonshot AI's Kimi K 2.6 and the previous DeepSeek V3.2.

The V4 Flash model, designed with smaller capacity, operates with 284 billion parameters. According to company experts, the new systems have almost entirely closed the gap with current leading international models in logical reasoning tests.

Data released by DeepSeek indicates that the V4 Pro Max model outperforms the latest systems from OpenAI and Google in logical tasks. In specialized programming competitions, both new models showed results on par with the world's most advanced technologies.

However, in tests measuring general knowledge, the Chinese models are still observed to be several months behind the most powerful US systems. The new models are priced much more affordably than other competitors in the market.

For instance, the cost of using the V4 Flash model is significantly lower than similar services from international tech giants. It is worth noting that this presentation took place during a complex period marked by US allegations of intellectual property theft against China and suspicions that DeepSeek copied other famous models.

Nevertheless, the new models are expected to further intensify competition in the field of artificial intelligence.