Chinese startup DeepSeek has unveiled its long‑awaited artificial intelligence model, DeepSeek‑V4, more than a year after its breakthrough R1 system rattled global markets.
The new model, released in two versions, promises ultra‑long context processing of up to one million words and drastically reduced computing costs. DeepSeek said the V4‑Pro contains 1.6 trillion parameters, while the lighter V4‑Flash offers a more economical option.
DeepSeek first rose to prominence in January 2025 with its R1 model, which matched the performance of U.S. rivals at a fraction of the cost. That release sparked a sell‑off in American tech stocks and was described as a “Sputnik moment” for the industry.
Analysts say the V4 launch underscores China’s growing role in artificial intelligence, though it is unlikely to trigger the same market frenzy as the R1. The company’s open‑source approach, combined with partnerships with Huawei to overcome U.S. chip export restrictions, highlights Beijing’s strategy to scale AI adoption despite geopolitical tensions.
DeepSeek’s release comes as global tech giants including OpenAI, Google, and Anthropic continue to push closed‑source models, while Meta and Microsoft restructure their workforces to invest more heavily in AI.
The rivalry between China and the United States in artificial intelligence is expected to intensify, with Washington accusing Chinese firms of intellectual property theft. For now, DeepSeek’s V4 marks another milestone in the race to make advanced AI cheaper and more accessible worldwide.
