China’s artificial intelligence-based startup DeepSeek, which shook the world markets last year, has released preview versions of its latest major update on 24th April, as the AI competition between China and the US intensifies.

V4 has been eagerly awaited by users seeking to explore the capabilities of DeepSeek compared to its U.S. counterparts, such as OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini. Anthropic and OpenAI have condemned DeepSeek for unfairly creating its tech based on their own.

The new model was expected to be available over a month earlier, at the beginning of the Lunar New Year, by some industry analysts.

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

V4 is an upgraded version of V3, an AI model that DeepSeek released at the end of 2024.

It was DeepSeek’s specialised “reasoning” AI model, R1, that caught markets off guard with its release in January 2025. DeepSeek argued that it was cheaper than OpenAI’s comparable model, and it served as a sign of China’s technological progress in its attempt to match the U.S.

In comparison to OpenAI’s GPT-5.2 model and Google’s Gemini 3.0-Pro, DeepSeek claims that the “V4 Pro Max” version has “superior performance” in terms of standard reasoning benchmarks. It is “marginally” inferior to GPT-5.4 and Gemini 3.1-Pro, it said.

When it comes to “agentic” abilities, the Chinese firm claimed the V4 “pro” model could outperform Claude Sonnet 4.5 and that it’s “close to the level of Claude Opus 4.5,” according to its own assessment.

On simpler agent tasks, DeepSeek claimed that the “flash” version of V4 functions at the same level as the “pro” version, and on the reasoning capability, the “flash” one is close to the “pro” version.

The benchmark results indicate that DeepSeek V4 will be a very strong contender against the U.S. competitors, Lian Jye Su, Omdia’s chief analyst, said.

Marina Zhang, associate professor at the University of Technology Sydney, described DeepSeek’s V4 launch as a “pivotal milestone for China’s AI industry”, particularly given the growing competition worldwide for the country to strive for self-reliance in key technologies.

DeepSeek provides a completely free web/mobile chatbot. Unlike the best of Anthropic, Google and OpenAI, it says its technology is “open source” in the sense that it will give developers access to be able to modify and build on its core technology.

Both the V4’s “pro” and “flash” versions have a 1 million token context window, a parameter of how much information an AI model can process and recall, and run on a more efficient basis, the startup said. This is a huge step up from the previous, as the V3 had a 128k context window.

A report from Microsoft in January showed that the use of DeepSeek has been gaining ground in many developing nations.

But some analysts are still sceptical. V4 is a “competent” follow-up, but isn’t quite as radical as R1, said Ivan Su, a Morningstar senior equity analyst.

Since R1’s release, domestic competition has greatly increased,” Su said. “Based on its own assessment, DeepSeek is quite comparable with most capabilities, on par with the US models, but independent assessments are required before anything definitive can be determined.”

In February, Anthropic was claiming that DeepSeek and two other Chinese AI labs are engaged in “industrial-scale campaigns” of “illicitly extracting Claude’s capabilities to boost their own models”. It said they did that using a technique called distillation that “involves training a less capable model on the outputs of a stronger one.” OpenAI has said the same in a letter to U.S. lawmakers.

This week, also, Michael Kratsios, Chief Science and Technology Adviser to US President Donald Trump, called out AI tech firms “domically based primarily in China” for “theft and selective appropriation of the fruits of American expertise and innovation.”

The Chinese embassy in Washington responded to the claims, calling them “unjustified suppression of the Chinese companies by the USA.”