DeepSeek Shocks OpenAI With AI For Less Money: How These AI Models Compare And Perform

Last Updated:January 28, 2025, 07:30 IST

DeepSeek has got everyone's attention in the US as the free app has topped the App Store charts in quick time.

DeepSeek R1 comes from China and it has got US worried, know why

OpenAI has been making waves since ChatGPT came to the market, forcing giants like Google and Apple to sit up and take notice. Now, these US-based companies are facing a major threat from China, in the form of the open source AI model from a startup called DeepSeek.

This company claims to have built its AI model using a different technology compared to its US rivals which has allowed them to spend less money to get similar results to the ChatGPT. So what is so special about the DeepSeek AI model, why has it become a big deal in a short time and why is the free app topping the App Store charts in the US? Here’s a detailed look at the new AI model in town that puts China back in the headlines.

Where It All Started

Before we get to the AI mechanics from the company, DeepSeek was launched in 2023 by Liang Wenfeng, who was basically running a hedge fund when he was 30 years old in 2015. Eight years later, Wenfeng had set up an army of PhDs from Chinese Universities which has thrown up an AI model that is based on a different technology from OpenAI that seems to be relying on less hardware to give the same results as the US giants.

DeepSeek AI: New AI Tech, Less Cost To Train

DeepSeek did the smart thing by predicting the future, and was able to buy around 10,000 NVIDIA GPUs before the US export ban came into effect. In total, the Chinese startup developed its AI model using 50,000 GPUs which is a fraction of the 500,000 GPUs that each of OpenAI, Google and other AI companies have deployed, as per reports.

But having all of this computing power would have been pointless if the results were not positive, which they sure were. The team of PhDs at DeepSeek have been paid handsomely, which their talent demanded, and one of the best on offer in the country.

But the biggest reason for DeepSeek AI to become a strong contender in the AI race is the ability to let the AI model learn from the data, and have a system in place which rewards the AI for giving the correct response. Compared to OpenAI, these tweaks have played a big role in expediting the AI evolution for the Chinese company.

DeepSeek vs OpenAI: Where It Differs

The detailed paper from DeepSeek AI clearly defines the strategy and technology that has given it a leg up over its biggest rival in the market.

The use of the Reinforcement Learning or RL has resulted in the DeepSeek R1 AI model to reason with its responses and deliver results much faster. The most important use of the RL technology is that DeepSeek was able to adapt and make AI do the heavy work with less money in the tank.

The OpenAI technology model called Supervised Fine-Tuning (SFT) follows the traditional channel, where you feed data into the AI model which gives it a first-hand understanding of the problems to solve. DeepSeek also has the open source bargain which makes it appealing for researchers, which explains its App Store rank shooting up in a short time.

The news about DeepSeek has wreaked havoc in the markets this week, and understandably so. After all, there you have OpenAI pitching for $500 billion in funds, and on the other side, DeepSeek is building AI at a much cheaper cost. But DeepSeek will have a cut off, and it will need more funds to run at this level and the open source model means the AI tech will gradually become part of other models as well.

News tech DeepSeek Shocks OpenAI With AI For Less Money: How These AI Models Compare And Perform