Chatgpt Vs Deepseek: The New Bane Of Openai's Existence
— written by Francis on Jan. 28, 2025, 8:04 p.m.
AI seems to be the benchmark of modern innovation when it comes to Tech companies. For a long time, OpenAI's ChatGPT has been the most superior language model, unrivaled by no other, not even Google's Gemini. However, a new rival from the East has risen. DeepSeek, a Chinese AI model has released its new open-source language model called, DeepSeek. I guess they ran out of names for their language model. This new language model has sent shivers down the spines of American Tech companies, with President Donald Trump calling it a "wake-up call" to American Tech companies.
But the question on everyone's mind is, which language model is the GOAT?
Comparisons to other language models
ChatGPT-4 and Deepseek have all proven to be superior to Google Gemini 1.5. Don't get me wrong, Gemini is a powerful AI, however, the 1.5 version works more like a search engine than a language model making it have an artificial tone. ChatGPT-4 on the other hand has a casual tone and occasional use of emojis which makes it feel less like an AI and more like a human... sounds weird saying that aloud. It's ability to adapt to each person also gives it a more realistic and human experience.
Deepseek on the other hand has a playful tone and uses emojis frequently making it feel interactive and lively. Not only this but if you enable its DeepThink ability, you'll get to see how it overthinks many things before giving a response which adds a layer of humor to it, just ask it to write a 100-character exact sentence and watch the chaos ensue 🌚🌚🌚, you'll thank me later.
So as much as I'd like to have a definitive winner here, it's a tie for this.
Realtime information
ChatGPT-4 is a beast when it comes to being up-to-date with ongoing events. This is due to its ability to access the internet and scrap website pages. For DeepSeek, it doesn't have access to the internet yet so it relies on its training data that runs up to July 2024. So for this, chatGPT has the upper hand.
Response speed and accuracy
When it comes to response speeds and accuracy, things don't look so good for ChatGPT when your free tokens run out and you are running on ChatGPT-4o mini. Let's look at the Kryptonite of language models, logic. This is a simple question that I asked DeepSeek, ChatGPT-4o and 4o-mini, and Gemini 1.5 and 2.0(the experimental version), "Alice has four brothers and she also has a sister. How many sisters does Alice's brother have?"
The answer of course is 2 sisters, however only ChatGPT-4o, DeepSeek, and Gemini 2.0 got it on the first try, and ChatGPT-4o mini got it after two tries. Gemini 1.5 never got it right even after asking the question 10 times. Ladies and Gentlemen, the receipts!
For accuracy ranking, I'll give this one to DeepSeek since it repeatedly answered the question correctly despite it being free and not able to access the internet. This is because it breaks down the question into a series of steps ensuring it answers accurately. If money wasn't a factor, it would be a tie between DeepSeek and ChatGPT-4o.
What of response time? Well, Gemini 1.5 responded in less than half a second, but with the wrong answer every time. ChatGPT-4o and 4o-mini still responded in under a second with better accuracy why DeepSeek's breakdown process meant it took around 2 seconds to respond to the riddle, but always accurately.
Conclusion
While ChatGPT-4o has established itself as a capable AI, DeepSeeks is closely behind it. When putting into consideration the capabilities of DeepSeek despite the fact it cannot access the internet, it is as capable as ChatGPT-4o and dare I say, probably would be more capable if it had access to the internet. OpenAI, watch out!
Comments