What is this buzz between OpenAI and DeepSeek?

How a Chinese Startup is Disrupting the AI Race and Challenging U.S. Tech Giants

DeepSeek, a relatively new player in the artificial intelligence space, has taken the industry by storm with its advanced language models, DeepSeek-V3 and DeepSeek-R1. It’s not just hype—DeepSeek recently surpassed OpenAI’s ChatGPT to become the most downloaded app on the U.S. Apple App Store.

This achievement couldn’t have come at a more pivotal moment. With Donald Trump back in the White House and U.S.-China tensions escalating over technology, DeepSeek’s rise has sparked debates about the future of AI dominance. As the U.S. tightens its grip on AI exports to China, DeepSeek’s momentum is challenging the long-standing dominance of American tech giants like OpenAI, Google, and Meta.

What is DeepSeek?

DeepSeek is a Chinese artificial intelligence startup founded in 2023 by billionaire Liang Wenfeng, who is also the co-founder of High-Flyer, a renowned quantitative hedge fund in China. Initially, High-Flyer focused on leveraging machine learning for stock trading, but in 2023, it spun off its AI research division to form DeepSeek. The startup is dedicated to developing cutting-edge AI models and aims to reach artificial general intelligence (AGI)—a level of AI capable of performing tasks equal to or surpassing human intellect.

DeepSeek gained international attention with the release of its open-source language models, including DeepSeek-V3 and DeepSeek-R1. These models are not only recognized for their advanced capabilities but also for their efficient, cost-effective development. For example, DeepSeek built its V3 model for under $6 million, a fraction of the cost incurred by Western counterparts like OpenAI and Meta.

The company’s innovation lies in its ability to optimize resources. Due to U.S. export restrictions on high-performance chips like Nvidia’s H100, DeepSeek relied on older and cheaper Nvidia H800 GPUs. This challenge led the startup to develop smarter, more energy-efficient algorithms, making it a formidable competitor in the global AI race.

DeepSeek’s models have earned praise for their performance in tasks such as text generation, reasoning, and answering complex questions. However, the platform does have some limitations, particularly around politically sensitive topics in China. Despite these constraints, DeepSeek is reshaping the AI landscape by challenging the dominance of U.S.-based tech giants like OpenAI, Google, and Meta, proving that ingenuity can thrive even in the face of resource constraints.

DeepSeek's R1 Model vs ChatGPT O1

Launched on January 21, 2025, DeepSeek's R1 model is an open-source alternative that emphasizes cost-effectiveness and accessibility. Its architecture utilizes a Mixture-of-Experts (MoE) approach, activating only a fraction of its parameters per token processed. This design allows R1 to maintain high performance while minimizing computational overhead. Notable features include:

• Mathematical Reasoning: Achieving impressive accuracy on benchmarks like MATH-500.
• Cost Efficiency: Pricing significantly lower than o1, making it attractive for developers and businesses.
• Transparent Reasoning Process: Users appreciate the model's ability to provide insights into its decision-making process.

Despite its strengths, R1 may experience slower processing speeds for certain tasks compared to o1.