DeepSeek V4 Preview Extends China Open-Source AI Reach
Chinese startup DeepSeek unveiled a preview of its V4 artificial intelligence model on Friday. The company plans to release the full version as open source. This move builds on its earlier success and aims to push China's presence further in global AI development.
DeepSeek first gained attention in January 2025. The firm claimed it developed an advanced AI system using far fewer resources on computer chips compared to U.S. competitors like OpenAI and Anthropic. That announcement created ripples across the tech sector.
DeepSeek's Industry Shake-Up
The January 2025 reveal kicked off what observers call China's "DeepSeek moment." This term captures the idea that Chinese AI companies now stand ready to display their skills on the world stage. DeepSeek made its model open source. That allowed anyone to use or alter it freely.
OpenAI and Anthropic, by comparison, maintain their top models as proprietary. Closed systems restrict access to protect business interests. DeepSeek's approach proved an open-source model could nearly match closed ones in performance. Costs played a key role too. Chinese developers showed they could build capable systems without massive spending on hardware.
Reporters Meaghan Tobin in Taipei, Taiwan, and Cade Metz in San Francisco covered the story. Their April 24, 2026, article, updated at 11:47 a.m. ET, highlights these shifts.
Growth of Chinese Open-Source Models
After DeepSeek's launch, other Chinese companies followed suit. They put out dozens of open-source AI models in the months ahead. By the end of 2025, these efforts claimed a large portion of worldwide AI usage. This trend marked a clear change in how AI technology spreads globally.
DeepSeek itself emerged as a notable player among Chinese AI startups. Focused on efficient model training, the firm leverages local advantages in hardware access and talent. Its models target practical uses like coding and reasoning, areas where performance matters most.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
V4's Strong Coding Performance
The V4 preview stands out for its code-writing abilities. This skill grows vital for top AI systems as they handle programming tasks. Tests by Vals AI, which monitors AI performance, show V4 beating every other open-source model in code generation.
Vals AI provides independent benchmarks. Their results confirm V4's edge in this area. While U.S. models from Anthropic and OpenAI lead overall in coding, Chinese systems like V4 close the distance steadily.
DeepSeek timed the V4 preview carefully. It came just days after Moonshot AI dropped its Kimi 2.6 model, also open source. Moonshot AI, another rising Chinese startup, focuses on versatile language models. Kimi series gained popularity for handling long contexts and complex queries.
Narrowing Gap with U.S. Leaders
Both V4 and Kimi 2.6 lag behind Anthropic's and OpenAI's best in coding prowess. However, the difference shrinks over time. This progress signals China's growing technical strength despite U.S. export controls on advanced chips.
OpenAI, founded in 2015, pioneered large language models with GPT series. Anthropic, started by ex-OpenAI staff, builds Claude models emphasizing safety. Both rely on huge compute investments. Chinese firms counter with optimized training methods and domestic hardware.
The open-source push benefits developers worldwide. It fosters innovation through community contributions. DeepSeek's V4 could accelerate code-related applications, from software development to automation scripts.
As Chinese models mature, they challenge the dominance of proprietary U.S. tech. The DeepSeek moment evolves into a sustained effort. Future releases will test if openness and efficiency can rival scale and secrecy.