Explained | What makes China’s open-source AI startup DeepSeek special

China-based open-source artificial intelligence startup DeepSeek has triggered a tech stock selloff globally with Indian IT firms HCL Technologies, Infosys, TCS and Tech Mahindra crashing by up to 5 per cent on Monday. Gyanendra Keshri delves into the new breakthrough AI tech and how it may challenge the dominance of American firms in the sector.

What is DeepSeek?

DeepSeek is an artificial intelligence research lab based in Hangzhou, China. Its AI model DeepSeek-R1, which was released in early January, has topped the iPhone download chart in the US, leaving behind the industry giant OpenAI’s ChatGPT. It is an open-source model. This means the developer community can inspect and improve the software.

Who owns DeepSeek?

DeepSeek is solely funded by Fire-Flyer, a Hangzhou-based hedge fund and AI firm. The company was founded in 2015 by Liang Wenfeng. However, the company claims that DeepSeek lab, which has released DeepSeek-R1 and its variants like DeepSeek-R1-Zero, operates independently. DeepSeek’s journey started in 2023 with the release of DeepSeek Coder, an open-source model designed for coding. This was followed by DeepSeek LLM and DeepSeek-V2.

Why has it grabbed global attention?

DeepSeek claims to offer equally efficient or even better technology intelligence at a fraction of cost. It also offers a counterpoint to the widespread belief that the future of AI will require ever-increasing amounts of power and energy to develop. According to a report by Epoch AI, DeepSeek’s R1 model required just one-tenth of the computing power used by Meta’s comparable Llama 3.1 model. “Jevons paradox strikes again! As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can't get enough of,” Microsoft CEO Satya Nadella wrote on X.

How are DeepSeek’s models different from OpenAI?

DeepSeek claims that unlike OpenAI’s ChatGPT its chatbots articulate its reasoning before delivering a response to a prompt. The company claims its R1 release offers performance on par with OpenAI’s latest and has granted license for individuals interested in developing chatbots using the technology to build on it. Advanced technical designs such as multi-head latent attention (MLA) and a mixture of experts made DeepSeek’s models more cost-effective. According to a post on DeepSeek's official WeChat account, DeepSeek-R1 is 20 to 50 times cheaper to use than OpenAI o1 model, depending on the task,

What does it mean for common users?

DeepSeek is likely to bring down the cost of artificial intelligence services significantly. Its advanced DeepSeek-R1 is free for users while similar chatbots like the OpenAI o1 and Claude Sonnet and Google Gemini, require a paid subscription. Moreover, DeepSeek doesn’t seem to have limits, while OpenAI, Claude and Google put restrictions on usage and only the older versions are available for free.