DeepSeek V4 Unveiled: China’s AI Upstart Challenges OpenAI and Google with Domestic Chip Breakthrough

HONG KONG — One year after the AI startup DeepSeek took the global technology industry by storm, the Hangzhou-based company has unveiled a preview version of its much-anticipated new model, V4. Released on Friday, the latest model promises to rival industry leaders from OpenAI, Anthropic, and Google by offering significant upgrades in reasoning and autonomous "agentic" capabilities, such as writing software code.

‎

DeepSeek became a prominent name in the global AI race following the 2025 release of its R1 model, which delivered high-end performance at a fraction of the cost of its American competitors. While that earlier success caused volatility in U.S. markets, analysts suggest that V4 represents a steady continuation of that competitive trend. Unlike many proprietary American models, DeepSeek V4 remains open source, a strategy that has allowed Chinese firms to rapidly scale their technology across various sectors like e-commerce and robotics.

‎

A key highlight of the V4 model is its reliance on domestic hardware. Due to U.S. export controls on high-end processors from Nvidia and AMD, DeepSeek partnered with Chinese tech giant Huawei to fulfill its computing needs. The model was developed using Huawei’s "Supernode" technology, which clusters "Ascend 950" chips to provide the necessary power for world-class AI training. Analysts from Counterpoint Research noted that this move allows AI systems to be built and deployed without total reliance on American hardware, potentially accelerating global AI development.

‎

In its own research findings, DeepSeek claimed that V4 possesses the best agentic coding capabilities among all open-source models. While the company acknowledged that the model still trails industry leaders like Google’s Gemini in broad world knowledge, it maintains that V4 achieves world-class reasoning performance.

‎

However, the startup’s rapid ascent continues to be met with skepticism from U.S. officials and competitors. Anthropic and OpenAI have previously accused DeepSeek of illegally "distilling" or extracting capabilities from their models. These concerns were echoed on Thursday by Michael Kratsios, the White House director of the office of science and technology policy, who accused foreign entities of conducting industrial-scale campaigns to copy frontier AI models developed by U.S. companies. DeepSeek has not yet commented on the specific allegations regarding the V4 model.

‎

Explore: NBCPalmSprings.com, where we are connecting the Valley.