China's DeepSeek launches an update of its AI model

DeepSeek, the Chinese artificial intelligence startup that shook up world markets last year, has launched preview versions of its latest major update.

DeepSeek’s V4 will be tested to see how it compares with U.S. competitors like OpenAI’s ChatGPT, Anthropic’s Claude and Google’s Gemini.

DeepSeek says the new V4 open-source model previews, which include “pro” and “flash” versions, have big improvements in knowledge, reasoning and in their agentic capabilities, or the ability to perform complex tasks and workflows autonomously.

V4 follows V3, an AI model that DeepSeek released in late 2024.

It launched its specialized “reasoning” AI model, called R1, in January 2025.

That showed China's progress in catching up with the U.S. in technological advancements.