What is Deepseek? Here’s what you need to know

Learn more about DeepSeek, the Chinese startup challenging the AI industry with open-source models that may rival OpenAI’s

The AI landscape is undergoing a seismic shift as DeepSeek, a Chinese AI company, challenges the established order. In a short time, it has disrupted the industry with its AI models, which can reportedly match those of giants like OpenAI in terms of performance and cost-efficiency.

DeepSeek's approach, prioritising both performance and accessibility, has forced a re-evaluation of the prevailing narrative: that powerful AI requires immense resources and remains the exclusive domain of tech giants. It is getting significant interest from the United States, forcing established players to rethink their strategies. Here are five things to know about the disruptive new presence.

1. DeepSeek is a Chinese startup

DeepSeek is a Chinese startup founded in 2023. It quickly gained attention for developing high-performing, open-source AI models. Unlike many AI models kept secret by large companies, DeepSeek's models are freely available for others to use and build upon, fostering a more collaborative and open approach to AI development. DeepSeek's approach disrupts the traditional AI landscape by making advanced AI models accessible and affordable.

2. DeepSeek just launched a reasoning model called R1

DeepSeek-R1 is a groundbreaking reasoning model that showcases a significant leap in AI capabilities, particularly in complex reasoning tasks. Reasoning models aim to mimic human-like cognitive abilities, such as deductive, inductive, and abductive reasoning. Trained using advanced techniques like reinforcement learning, R1 excels in mathematics, coding, and problem-solving, often surpassing the performance of other leading AI models.

3. R1 can possibly rival OpenAI’s performance

DeepSeek-R1 showcases remarkable innovation by being on par with OpenAI's o1 in tasks like reasoning, coding, and math—while operating at a fraction of the cost. By leveraging pure reinforcement learning and streamlined training techniques, DeepSeek has drastically reduced computational expenses, cutting costs by as much as 95 per cent.

Jack Clark, co-founder of AI safety and research company Anthropic, wrote in his newsletter: “R1 is significant because it broadly matches OpenAI's o1 model on a range of reasoning tasks and challenges the notion that Western AI companies hold a significant lead over Chinese ones.”

4. DeepSeek R1 is open-source

Notably, R1 is open-source (i.e. free), which means the technology is accessible to anyone, from researchers and developers to AI hobbyists. By making the model's code and architecture publicly available, DeepSeek encourages experimentation, collaboration, and rapid innovation within the AI community.

Yann LeCun, Meta's chief AI scientist said in a post on Threads, “It's not that China's AI is ‘surpassing the US,’ but rather that 'open source models are surpassing proprietary ones.”

5. In a short time, DeepSeek R1 has disrupted the US tech market

DeepSeek's R1 has triggered a sell-off on AI-related stocks in the US tech market, triggering a sell-off in AI-related stocks. The release of the resource-efficient AI model has raised concerns about the future demand from major players, potentially impacting their revenue and stock prices. The AI model's open source also challenges the dominance of US tech giants. Adding to the turmoil, DeepSeek has faced large-scale cyberattacks, temporarily limiting user registrations and highlighting the security vulnerabilities associated with advanced AI systems. This has thrown US tech stocks into a bit of a frenzy, with investors trying to figure out how R1 will shake things up and what it means for the future of AI.

DeepSeek AI is contributing to the evolving field of AI by developing accessible and potentially impactful models.

Nvidia AI researcher Jim Fan agreed, posting on X: “We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive—truly open, frontier research that empowers all.”

While the full extent of their contributions remains to be seen, DeepSeek’s work highlights the ongoing exploration of how AI is still anyone’s game.

NOW READ:

Lessons for millions: 5 edtech startups from Asia to watch

This Chinese unicorn company uses AI to help Unilever and L’Oreal with their content

From the singularity to AI agents: Sam Altman on how AI will be in 2025 and beyond

Tatler Asia

5 things to know about AI industry disruptor DeepSeek

1. DeepSeek is a Chinese startup

2. DeepSeek just launched a reasoning model called R1

3. R1 can possibly rival OpenAI’s performance

4. DeepSeek R1 is open-source

5. In a short time, DeepSeek R1 has disrupted the US tech market