According to a story from The Information, Chinese AI startup DeepSeek intends to introduce its next-generation V4 model in mid-February as it expands into developing countries in spite of security concerns raised by Western governments.
According to The Information, which cited persons with knowledge of the situation, internal tests conducted by DeepSeek staff indicate that V4 may surpass competitors such as Anthropic’s Claude and OpenAI’s GPT series in coding jobs. Additionally, the model has made strides in handling incredibly lengthy coding prompts, which could provide professionals working on intricate software projects with a new edge.
Increasing Uptake In Developing Nations
DeepSeek has gained popularity in underdeveloped countries, according to a report published Thursday by Microsoft’s AI for Good Lab. The platform’s free and open-source methodology lowers obstacles in price-sensitive regions. According to the survey, 16.3% of people worldwide adopted generative AI technologies in the three months leading up to December, up from 15.1% in the preceding quarter.
According to the Microsoft analysis, DeepSeek has an 89% market share in China, followed by Belarus at 56%, Cuba at 49%, and Russia at 43%. The platform’s market share was between 11% and 14% in a number of African nations, including Ethiopia, Zimbabwe, Uganda, and Niger, while it was between 23% and 25% in Syria and Iran, respectively.
“DeepSeek’s rise shows that global AI adoption is shaped as much by access and availability as by model quality,” the Microsoft paper said.
Chief data scientist Juan Lavista Ferres of Microsoft’s AI for Good Lab pointed out that although DeepSeek excels in coding and math, it behaves differently on political subjects than US-based models. “We have observed that for certain type of questions, of course, they follow the same type of access to the internet that China has,” according to him.
Increased Security Scrutiny
The startup from Hangzhou, which unveiled its R1 reasoning model in January 2025 and V3 in December 2024, has come under fire for its security and privacy policies in a number of nations. Citing national security threats and worries over data storage in China, Australia, South Korea, Taiwan, and Italy have limited or outright prohibited DeepSeek on government equipment.
According to the Microsoft analysis, DeepSeek adoption increased in regions where foreign tech access is restricted or where US services are restricted, but it remained low in North America and Europe.
A version of DeepSeek’s V3 model, which was published in December 2024, is the V4 model. Requests for comment from DeepSeek were not immediately answered, and Reuters was unable to confirm The Information’s report.