DeepSeek Releases V4 AI Model, Claims Breakthrough in Cost Efficiency and Context Length

HANGZHOU — Chinese AI startup DeepSeek has introduced its latest artificial intelligence system, DeepSeek-V4, claiming significant reductions in computing and memory costs while expanding model capability to process up to one million words of context.

The company announced the release on Friday through WeChat and X, describing the model as “world-leading” and highlighting its improved efficiency compared to earlier versions.

DeepSeek said the system’s extended context capacity “(achieves) leadership in both domestic and open-source fields across agent capabilities, world knowledge, and reasoning performance.”

A preview version of the open-source model is now available to developers and researchers.

The launch comes at a time of heightened competition in the global AI industry, particularly between China and the United States. On Thursday, the White House accused Chinese-linked entities of running what it called large-scale efforts to steal American AI technology.

DeepSeek first emerged as a major player in January last year with its R1 reasoning model, which powered a chatbot that matched leading US AI systems while reportedly requiring significantly lower computational resources.

Experts say the new release could reshape AI development economics.

“This addresses the long-standing issues of slower performance and higher costs associated with long context lengths, marking a genuine inflection point for the industry,” said Zhang Yi, founder of iiMedia research firm.

He added that broader adoption could transform how AI is used commercially. “For end users, this will bring widespread, accessible benefits. If ultra-long context support becomes standard, long-text processing is expected to move beyond high-end research labs and enter mainstream commercial applications,” he said.

DeepSeek said the model is available in two variants: DeepSeek-V4-Pro and DeepSeek-V4-Flash, with the latter designed as a lighter and more cost-efficient option.

According to the company, V4-Pro contains 1.6 trillion parameters, while V4-Flash has 284 billion parameters that enhance model decision-making capabilities.

The model has also been optimised for integration with AI agent platforms including Claude Code, OpenClaw, OpenCode, and CodeBuddy.

“In world knowledge benchmarks, DeepSeek-V4-Pro significantly leads other open-source models and is only slightly outperformed by the top-tier closed-source model, (Google’s) Gemini-Pro-3.1,” the company said.

DeepSeek’s earlier breakthrough sparked what analysts described as a “DeepSeek shock,” triggering a global sell-off in AI-related stocks and being labelled a “Sputnik moment” for the industry.

However, the company has faced ongoing scrutiny over data privacy and censorship concerns, with reports that its chatbot avoids politically sensitive subjects such as the 1989 Tiananmen Square crackdown.

Despite these concerns, DeepSeek’s tools are widely used in China across municipal governments, healthcare institutions, financial services, and private enterprises, supported by its open-source model strategy.

The announcement also comes amid broader restructuring in the global tech sector, with Meta planning staff reductions and reports suggesting Microsoft is also considering workforce cuts while increasing investment in artificial intelligence.

Africans Angle News