DeepSeek: A Complete Guide to the AI Chatbot App

This week, Chinese AI lab DeepSeek gained widespread attention as its chatbot app soared to the top of both the Apple App Store and Google Play charts. Trained using compute-efficient techniques, DeepSeek’s AI models have sparked debate among Wall Street analysts and technologists about the U.S.’s ability to maintain its AI leadership and the long-term demand for AI chips.

But what are DeepSeek’s origins, and how did it achieve global recognition so rapidly?

DeepSeek is backed by High-Flyer Capital Management, a Chinese quantitative hedge fund that leverages AI for trading strategies.

AI enthusiast Liang Wenfeng co-founded High-Flyer in 2015. While studying at Zhejiang University, he developed an interest in trading and later launched High-Flyer Capital Management in 2019 as a hedge fund specializing in AI-driven algorithms.

From High-Flyer’s AI Lab to Independent Company

In 2023, High-Flyer established DeepSeek as a separate AI research lab, distinct from its financial operations. With investment from High-Flyer, the lab later spun off into an independent company, also named DeepSeek.

From the start, DeepSeek built its own data center clusters for training AI models. However, like other Chinese AI firms, it has faced challenges due to U.S. export restrictions on hardware. To train one of its latest models, the company had to rely on Nvidia H800 chips—a less powerful alternative to the H100 chips available to U.S. firms.

DeepSeek’s technical team is reportedly composed of mostly young professionals. The company actively recruits AI researchers with doctorates from leading Chinese universities and also brings in individuals without computer science backgrounds to enhance its technology’s understanding of diverse subjects, according to The New York Times.

DeepSeek introduced its first models—DeepSeek Coder, DeepSeek LLM, and DeepSeek Chat—in November 2023. However, the AI industry truly took notice in the spring of 2024 when the company launched its next-generation DeepSeek-V2 model family.

DeepSeek-V2, a versatile system capable of analyzing both text and images, performed strongly in AI benchmarks while being significantly more cost-efficient than competing models. Its efficiency prompted domestic rivals like ByteDance and Alibaba to lower prices on some AI models and make others free.

The release of DeepSeek-V3 in December 2024 further cemented the company’s reputation. According to internal benchmark tests, DeepSeek-V3 outperformed both open-source models like Meta’s Llama and proprietary models such as OpenAI’s GPT-4o.

DeepSeek R1: A Self-Checking AI for Advanced Reasoning

Equally noteworthy is DeepSeek’s R1 “reasoning” model, launched in January 2025. DeepSeek claims R1 matches OpenAI’s o1 model on key benchmarks. Unlike traditional models, R1 can fact-check itself, reducing errors that commonly affect AI systems. While reasoning models take longer—sometimes seconds to minutes—to generate responses, they tend to be more reliable, particularly in fields like physics, science, and math.

However, as a Chinese-developed AI, DeepSeek’s models must comply with regulatory standards set by China’s internet authority, which ensures content aligns with “core socialist values.” For instance, in DeepSeek’s chatbot app, R1 avoids answering questions about Tiananmen Square or Taiwan’s political status.

DeepSeek’s business model remains unclear, as it prices products below market rates and offers some for free. The company claims cost-efficiency allows this, though some experts dispute it.

Despite uncertainties, developers have embraced DeepSeek’s models, which, while not fully open source, are available for commercial use. On Hugging Face, over 500 derivative models of R1 have been created, totaling 2.5 million downloads.

DeepSeek’s rise has been both praised and criticized. Its impact contributed to an 18% drop in Nvidia’s stock and drew a response from OpenAI’s CEO. Microsoft integrated DeepSeek into Azure AI Foundry, while Meta and Nvidia acknowledged its growing influence.

However, several companies and governments, including South Korea and New York state, have banned DeepSeek. Reports suggest the U.S. may follow due to concerns over foreign influence.

While DeepSeek is expected to refine its models, its future remains uncertain amid regulatory scrutiny.

Read the original article on: TechCrunch

Read more: Nvidia CEO Jensen Huang Dismisses DeepSeek as Sales Continue to Surge