SPONSORED
Elevate Magazine
February 2, 2025

Everything You Need to Know About DeepSeek AI

everything you need to know about deepseek ai
Photo source: CNET

DeepSeek has recently captured the world’s attention. It has made headlines this week as its chatbot application surged to the top of both the Apple App Store and Google Play charts. 

The innovative AI models developed by DeepSeek have prompted Wall Street analysts and technology experts to reconsider the United States’ position in the AI sector and raise questions about the sustainability of demand for AI chips.

Origins of DeepSeek

DeepSeek is supported by High-Flyer Capital Management, a quantitative hedge fund based in China that leverages AI for trading strategies. Co-founded by Liang Wenfeng in 2015, High-Flyer began as a hedge fund in 2019 with a focus on creating and implementing AI algorithms. In 2023, High-Flyer initiated DeepSeek as an independent lab that aimed to explore AI technologies apart from its financial operations, eventually leading to the establishment of the company known as DeepSeek.

From its inception, DeepSeek has constructed its own data centre clusters for training models. However, like many Chinese AI firms, it has faced challenges due to U.S. export restrictions on hardware. To develop one of its latest models, DeepSeek had to utilise Nvidia H800 chips, which are less powerful than the H100 chips available to U.S. companies.

The technical workforce at DeepSeek is reportedly youthful, with a strong emphasis on recruiting PhD-level AI researchers from leading Chinese universities. Additionally, the company employs individuals without formal computer science training to broaden its understanding across various fields.

DeepSeek’s Advanced Models

In November 2023, DeepSeek launched its initial models—DeepSeek Coder, DeepSeek LLM, and DeepSeek Chat. However, it was the introduction of the next-generation DeepSeek-V2 models in spring 2024 that garnered significant attention within the AI community.

This general-purpose system for analysing text and images excelled in various benchmarks while being more cost-effective than similar models, compelling competitors like ByteDance and Alibaba to lower their pricing or offer some services for free.

The subsequent release of DeepSeek-V3 in December 2024 further boosted the company’s reputation. Internal benchmarking indicated that DeepSeek V3 outperformed both freely available models like Meta’s Llama and proprietary models accessible only via API, such as OpenAI’s GPT-4o.

DeepSeek also introduced its R1 reasoning model in January 2025, claiming that it matches OpenAI’s o1 model on key performance indicators. As a reasoning model, R1 is capable of self-verification, which mitigates common errors encountered by other models. Although reasoning models generally require additional time—often seconds to minutes longer—to generate answers compared to standard models, they tend to deliver more accurate results in complex domains like physics and mathematics.

However, there are limitations associated with R1 and other DeepSeek models. As products developed in China, they must comply with regulations set by China’s internet authority to ensure that their outputs reflect “core socialist values.”

A Unique Business Strategy

The company offers its products and services at significantly lower prices than competitors and even provides some for free. According to DeepSeek, advancements in efficiency have allowed it to maintain this aggressive pricing strategy. However, some industry experts question the accuracy of these claims.

Despite this uncertainty, developers have embraced DeepSeek’s offerings. While not open source in the traditional sense, its models are accessible under permissive licenses that permit commercial use.

DeepSeek’s rise against larger competitors has been characterised as “disruptive” and “over-hyped.” The company’s success has contributed to a notable decline in Nvidia’s stock price by 18% on a single day and prompted a public response from OpenAI CEO Sam Altman.

Microsoft recently announced that DeepSeek is now available through its Azure AI Foundry service—a platform designed to consolidate various AI services for enterprises. When questioned about DeepSeek’s influence on Meta’s AI expenditures during a recent earnings call, CEO Mark Zuckerberg emphasised that investment in AI infrastructure would remain a “strategic advantage” for Meta.