BEIJING: Chinese artificial intelligence company DeepSeek announced the launch of its latest AI model on Friday, more than a year after gaining global attention with its affordable reasoning model comparable to U.S. counterparts.
The new model, named DeepSeek-V4, boasts an “ultra-long context window” of one million words, according to the company’s social media posts on WeChat and X (formerly Twitter). The company described it as a “cost-effective” solution in its separate announcement.
The rollout coincides with reports of major tech firms like Meta planning significant staff reductions up to ten percent as they seek increased productivity and continue to invest heavily in AI development. Microsoft is also reportedly considering workforce cuts.
DeepSeek-V4’s extended context capability allows it to process vast amounts of input, enhancing its performance in tasks involving reasoning, knowledge, and agent functions. The company claims the model leads both domestically and in open-source AI sectors in these areas.
A “preview” version of the open-source model is now accessible to developers. The model is available in two variants: DeepSeek-V4-Pro and DeepSeek-V4-Flash. The latter is designed to be more efficient and affordable, featuring fewer parameters.DeepSeek-V4-Pro is equipped with 1.6 trillion parameters, while the Flash version contains 284 billion, both aimed at improving decision-making accuracy. The models have been optimized for integration with popular AI agent platforms such as Claude Code, OpenClaw, OpenCode, and CodeBuddy.
In world knowledge benchmarks, DeepSeek-V4-Pro demonstrates significant leadership among open-source models, trailing only slightly behind Google’s top-tier Gemini-Pro-3.1, a closed-source model.DeepSeek made headlines in January last year with its generative AI chatbot, powered by the R1 reasoning model, which challenged the dominance of U.S. AI firms. Dubbed the “DeepSeek shock,” the launch led to a sharp decline in AI-related stock prices and prompted strategic reassessments within the industry described by some as a “Sputnik moment.”
The chatbot matched the performance of ChatGPT and other leading American models but required substantially less computing power to develop. Its popularity, however, raised concerns about data privacy and censorship, as it often refused to respond to sensitive topics such as the Tiananmen Square incident.
Within China, DeepSeek’s AI solutions have been widely adopted across municipal governments, healthcare providers, financial institutions, and various industries. The company’s open-source approach—making its models’ inner workings public stands in contrast to the proprietary systems offered by OpenAI and other Western firms.“China’s large AI models are leading the development of the global open-source AI ecosystem,” Chinese Premier Li Qiang remarked at a recent high-level gathering.
The AI race has intensified the rivalry between China and the United States. Recently, the White House accused Chinese entities of large-scale efforts to steal U.S. AI technology, with officials warning of possible actions to safeguard American innovation.
