BEIJING, China – Web Desk: Chinese artificial intelligence company DeepSeek has unveiled its latest AI chatbot models, intensifying competition with US-based rivals such as OpenAI and Google.
The company, which sent shockwaves through the global tech industry in January 2025 with the release of its R1 chatbot, has now introduced preview versions of DeepSeek V4 Pro and DeepSeek V4 Flash on April 24. DeepSeek had previously claimed to have developed an AI model comparable to OpenAI’s ChatGPT for under $6 million, causing significant financial ripples across American AI firms.
According to a social media post by the company, the new models are open-source, allowing developers to freely use and modify the source code. While DeepSeek V4 Pro reportedly outperforms other open models in mathematics and coding, it trails slightly behind GPT-5.4 (OpenAI) and Gemini 3.1 Pro (Google) in overall performance, placing the company about three to six months behind its US competitors. The Flash variant offers faster response times at a significantly lower cost.
DeepSeek was founded in May 2023 by Liang Wenfeng in Guangdong Province. Unlike profit-driven AI firms, DeepSeek operates purely as a research-focused entity. Liang, who previously founded the hedge fund High-Flyer Capital, financed DeepSeek through personal capital. In 2021, his assets exceeded 100 billion Chinese yuan. Reports suggest Liang had acquired over 10,000 Nvidia graphics processing units before US export restrictions tightened under former President Joe Biden in 2022, targeting powerful H100 and later H800 chips.
Liang has stated that the biggest challenge was never capital but securing powerful chips amid US sanctions. He prioritizes local talent for the company’s Guangzhou campus.
