DeepSeek vs ChatGPT – How do they compare?

By
Surakshya Acharya
-
DeepSeek vs ChatGPT
Asus Ad

When ChatGPT was first introduced, it became an instant favorite to everyone. But now, there’s a new player in town: DeepSeek. This open-source, completely free AI model from China is not only giving ChatGPT tough competition but is also proving that the future of AI might not be as expensive as we thought. Let’s get into the depth of DeepSeek vs ChatGPT.

DeepSeek vs. ChatGPT

When ChatGPT hit the stage, it was like the AI world had found its rockstar. Developed by OpenAI, ChatGPT quickly became the go-to tool for everything from writing essays to debugging code and even crafting dad jokes (hahaha!). Its ability to understand and generate human-like text made it a global sensation, and for good reason.

But here’s the catch: while ChatGPT’s early versions were free, the latest and greatest iterations like GPT-4 (and newer) come with a price tag. Sure, you get what you pay for, but not everyone is willing to spend money for AI-powered small talk.

Enter DeepSeek, China’s answer to the AI arms race. This open-source, completely free AI model is not just keeping up with ChatGPT it’s giving it serious head-to-head challenges.

DeepSeek R1

Why DeepSeek stands out

DeepSeek is a powerful AI tool that’s been making big headlines lately. It’s open-source, meaning developers worldwide can tinker with it, improve it, and make it their own. And the best part? It’s absolutely free. No subscriptions, no hidden fees, and no “premium features” locked behind a paywall. Recently it was revealed that the Chinese company had created Deepseek for around 5 million dollars, which created a massive upheaval in the US market, particularly that of the chip company like NVIDIA, who lost more than 600 Billion dollars in a single day.

But DeepSeek isn’t just about being free it’s also a powerhouse in terms of performance. In several key areas, it’s already surpassing ChatGPT. For instance, DeepSeek excels in handling complex mathematical problems, is incredibly efficient at processing and summarizing large volumes of text, let’s not forget its multilingual capabilities, which are giving ChatGPT a tough time in the global market. I’ve written a separate article where you can see how DeepSeek competes equally and even outperforms tools like ChatGPT-4 and Claude 3.5.

deepseek benchmark score

On the downside, ChatGPT provides fair answers to any question, no matter the country of origin, however, DeepSeek refrains from answering questions related to China’s censored news or sensitive topics.

How do they compare?

Pricing

DeepSeek R1 stands out for its affordability, you can use it for free, while OpenAI charges $20 for ChatGPT o1. For developers, its API is 27x cheaper than ChatGPT o1. Additionally, the DeepSeek team has open-sourced its RL method for test-time computing.

1V1

I compared them in multiple categories, and here’s how it went:

Q1. Reasoning 

“if a train leaves Station A at 10:01 AM, traveling at 60 km/h, and another train leaves Station B at 11:00 AM, traveling at 80 km/h toward Station A, at what time will they meet if the distance between the stations is 300 km?”

Both of the models were able to reason out pretty well with the same answer.  However, my view is that the chatgpt had a much clearer explanation that could be well laid out to a layman.

Q2. Creative writing

“Write a short story (100 words) about an anti-hero who’s alienated from his society”

So talking on “creative” writing, I wasn’t particularly impressed by either of the models. However, Deepseek has the edge as it attempts to amalgamate alienation with a more pronounced set (creating a mythical fiction-like background), compared to ChatGPT who seems to bring out the most cliched definition of alienation in existence. Above we can see, how Deepseek uses the role of an anti-hero as a universally applicable being and not just as limited to a modern individual who’s “anti-establishment”. The neon-lit set also explains how ChatGPT just vomits the typical/cliched view of lighting and staging as something very mechanical and having no individuality of its own.

3. Knowledge/ Essay Writing

“Grierson, Flaherty and several other documentary pioneers were explicitly influenced by the strategies of Russian cinema, particularly their use of montage, yet these strategies have for the most part been rejected by more contemporary documentary film practitioners. Examine why (or if!) you think this to be the case, citing examples. [ In 250 words]”

Regarding knowledge/history-based questions, I think both of them did pretty well. But my view is that ChatGPT gave a better answer, as it also countered the fact that the strategies of Russian cinema are still in action and not rejected in their entirety.

Conclusion: ChatGPT vs Deepseek

In my opinion, both of them have strengths and weaknesses. Aside from its political censorship of anything related to China, I think it does a commendable job compared to its competitors. However, because of the low operating budget, Deepseek users can face problems with traffic as more and more people are trying to access it.

A few more words

While the U.S. has been leading the AI race for years, China is catching up fast. DeepSeek is just one example of how Chinese AI is becoming an alarming competitor. Looking ahead, the possibilities are endless. And with models like DeepSeek leading the charge, the future of AI might just have a “Made in China” label.