DeepSeek: ChatGPT killer or just another hype train? We compare it against ChatGPT and Gemini

DeepSeek, the latest AI chatbot from China, has, in the past week or so, become one of the most intensely discussed topics online. What brought it to fame was its V3 large language models (LLM), which have capabilities similar to top AI chatbots such as ChatGPT but with much lower development costs while using less powerful Nvidia H800 chips.

However, it was later reported that DeepSeek is only using the Nvidia chips mainly to train its AI chatbot. Tech tipsters are claiming that Huawei’s chipsets are the main chipsets running the show behind the curtain.

The debut of DeepSeek triggered a panic reaction in the US stock market, causing Nvidia to lose around USD600 billion in market cap, the biggest one-day loss in US history. Countries such as Australia and South Korea have even banned it from being used by government employees. The Chinese AI chatbot’s impact is so large that US President Donald Trump depicted it as a “wake-up call” for American tech companies.

But is DeepSeek really as great as what people claim it to be, and is this the end for American AI companies? We conducted a mini-comparison between DeepSeek using its V3 and R1 LLMs, and America’s top two AI chatbots, OpenAI’s ChatGPT 4o and Google’s Gemini Advanced.

To test their capabilities, we gave the chatbot tasks such as giving feedback on our article piece, summarising an article based on a link, and testing them on their knowledge towards recent events. Watch the video above to see what we’ve discovered.