Close

China’s DeepSeek releases long-awaited new AI model

BEIJING: Chinese startup DeepSeek has unveiled a new artificial intelligence model, more than a year after it gained global attention with a low-cost reasoning system that rivalled leading US AI tools.

In a statement posted on WeChat and X, the company said DeepSeek-V4 “features an ultra-long context of one million words” and described the model as “cost-effective”.

The release comes at a time when major US tech firms, including Meta and Microsoft, are reportedly restructuring operations while increasing investment in artificial intelligence.

DeepSeek said the V4 model delivers strong performance across “agent capabilities, world knowledge, and reasoning”, and is now available in a preview version as open source.

The model is offered in two variants: DeepSeek-V4-Pro and DeepSeek-V4-Flash. The Flash version is positioned as a more efficient option due to its smaller parameter size.

According to the company, V4-Pro contains 1.6 trillion parameters, while V4-Flash has 284 billion parameters, which help determine how the model processes and refines outputs.

DeepSeek also said the model has been optimised for use with several AI agent tools and coding platforms, including Claude Code, OpenClaw, OpenCode, and CodeBuddy.

The company claimed that in world knowledge benchmarks, V4-Pro outperforms other open-source models and comes close to top closed-source systems such as Google’s Gemini-Pro-3.1.

DeepSeek first drew international attention in January last year with its R1 chatbot model, which delivered strong performance at lower computational cost, sparking what analysts called a “DeepSeek shock” across global tech markets.

Its rapid rise raised both interest and concerns over data privacy, censorship, and geopolitical competition in AI development between the United States and China.

DeepSeek’s tools have since been adopted across various sectors in China, including government, healthcare, and finance, supported by its open-source approach.

The latest release further intensifies the ongoing AI competition between Beijing and Washington, with both sides accusing each other of industrial and technological espionage in the sector.

Leave a Reply

Your email address will not be published. Required fields are marked *

Leave a comment
scroll to top