DeepSeek Looks To Launch Its New AI Model Earlier Than Expected

According to the media report, DeepSeek is planning to accelerate the release of its next-generation AI model, R2.

Sources familiar with the matter revealed that R2 was originally scheduled for release in early May, but the company now aims to launch it sooner. DeepSeek stated that the new model will feature improvements in programming capabilities and multilingual reasoning.

Vijayasimha Alilughatta, Chief Operating Officer of Indian tech service provider Zensar, believes that "The launch of DeepSeek's R2 model could be a pivotal moment in the AI industry." He added that its cost-effective approach to AI model development "would likely spur companies worldwide to accelerate their own efforts ... breaking the stranglehold of the few dominant players in the field."

Analysts attribute DeepSeek's success to over a decade of substantial investment in research and computing power by its backer, quantitative hedge fund "High-Flyer". Between 2020 and 2021, Huanfang invested 1.2 billion yuan to establish two AI supercomputing clusters, which include approximately 10,000 Nvidia A100 chips.

Former employees of DeepSeek noted that the company's success is largely due to founder Liang Wenfeng's focus on more cost-effective AI architectures. DeepSeek has adopted technologies such as Mixture of Experts (MoE) and Multi-Head Latent Attention (MLA), significantly reducing computational costs.

The report also highlighted DeepSeek's flat management structure, which fosters a collaborative atmosphere. Founder Liang Wenfeng, described by former colleagues as low-key and introverted, prefers to dive into technical details alongside younger employees.

A former employee quoted in the report said, Liang gave us control and treated us as experts. He constantly asked questions and learned alongside us.

In general, DeepSeek looks to challenge industry giants and inspire global enterprises to accelerate their AI efforts once again, and this time, it may aim to completely shape the future of the AI industry.