Loading [MathJax]/extensions/tex2jax.js

DeepSeek. ChatGPT but for a fraction of the cost?

DeepSeek vs. ChatGPT

Not long ago, DeepSeek, a Chinese tech startup, launched R1, a model claiming to compete with ChatGPT’s top-of-the-line o1 model, taking over headlines and tanking the American stock market. DeepSeek released its first AI model in the final quarter of 2023 with the Coder and continued to make efficient models until the massive breakthrough with R1, a reasoning model similar to o1 by OpenAI. These models are specifically designed to complete challenging multistep problems and are heavily advertised for STEM issues, including scientific inquiry, coding, and mathematical challenges. Due to minimal access to Nvidia’s A100 and H100 chips for AI development, as a result of the bans on shipping these American graphics cards to China in 2022, the Chinese researchers under Liang Wenfeng had no choice but to create a model as capable as o1 but with significantly fewer resources.

And they did it.

DeepSeek R1 was developed with just $6 million, shaking the American economy. Comparing this to OpenAI’s $100 million budget for ChatGPT-4o, the Chinese startup’s feat made OpenAI’s efforts seem wasteful. Knowing the controversial influence of the Chinese government, several countries have already banned DeepSeek; however, this doesn’t dismiss the point that OpenAI does not have a monopoly over AI models. DeepSeek demonstrated that AI development is not all about more resources, memory, and GPU power, damaging the reputation of Nvidia (17% decline from $3.5T value) and bringing down NASDAQ by 3%. 

President Trump emphasized that this is a wake-up call to American companies. Ultimately, o1 by OpenAI still maintains an edge over DeepSeek R1 in terms of creativity performance, while they remain even in reasoning, showcasing the value of additional resources. Meanwhile, China has imposed heavy censorship over DeepSeek R1, risking bias in the model’s learning over time. Competition brings out the best, so the next generation of OpenAI’s reasoning models will likely showcase incredible advancements, perhaps with a greater focus on cost efficiency. 

About the Author

Yash Yardi
I am the Opinions Section Editor on the Acronym! I am currently in 1504 A wing. I have several hobbies in STEM including coding, VEX robotics, and writing for an AI Newsletter. The Acronym is one of my favorite clubs at IMSA for its fantastic working environment, prominence, and learning opportunities!

Be the first to comment on "DeepSeek. ChatGPT but for a fraction of the cost?"

Leave a comment

Your email address will not be published.


*