I Tried This New Chinese-Developed Super-Powerful AI Model

Artificial intelligence is advancing rapidly, and China-developed chatbot DeepSeek AI is one of the latest powerful technologies making waves. There’s a lot to like about DeepSeek’s quick and detailed answers, but it does have some shortcomings that mean it’s not the AI ​​tool you’re looking for.

What is DeepSik?

DeepSik is a powerful new artificial intelligence tool developed by a Chinese technology company. Its flagship model, DeepSeek-V3, uses a unique Mixture-of-Experts (MoE) architecture. Think of it as a “team” of specialized artificial intelligence systems in which only the most suitable experts are “activated” to solve specific problems, where each “expert” is a specialized neural network.

DeepSeek claims that its LLM V3 was trained on a whopping 14.8 trillion tokens, with one million tokens equivalent to approximately 750,000 words. DeepSeek V3 is also a huge LLM overall, with 671 billion parameters in the MoE architecture, with 37 billion of these parameters activated per token. This means that behind every invitation there is a huge amount of processing power, ensuring faster and more efficient work.

Moreover, it is an open source artificial intelligence modelThis means that everyone can research, test and develop it for free. This sets it apart from competitors who keep their models tightly behind a paywall.

DeepSeek Challenge

To see how DeepSik-V3 works, I tested it on three practical tasks: creative writing, programming assistance, and productivity-related work. Its performance demonstrated both versatility and raw power, with some minor hiccups.

Task 1: Writing a description of a fictional character

I asked DeepSeek to write a detailed description of a fantasy character: a rebel queen who leads the resistance to an evil empire. The result was bright and interesting.

Task 2: Debugging a JavaScript Function

I tested DeepSeek’s programming skills by giving it a buggy JavaScript function that calculates the factorial of a number. Here is the error code I provided:

        ​​​​function factorial(n) {  
   if (n = 1) {  
       return 1;  
   }  
   return n * factorial(n - 1);  
} 
​
    

DeepSeek immediately discovered the problem: the single equals sign (=) in the condition was causing a logic error. He provided the corrected code and also explained the problem:

This fix worked great and the explanation was clear enough for a JavaScript newbie to understand.

Task 3: Productivity – Creating a Meeting Agenda

For a performance test, I asked DeepSeek to create a brief agenda for a project team meeting about a new product launch.

DeepSeek excels in creative writing, coding, and performance-related tasks, delivering impeccable results across the board. As with any artificial intelligence tool, it is important to check the accuracy of the results, especially when accuracy matters.

The good, the bad and the amazing

After testing DeepSeek, several strengths and weaknesses emerged. It also surprised me in a few unexpected ways.

DeepSeek easily handles a wide range of tasks. From producing vibrant, creative writing to debugging complex code, it has demonstrated flexibility that rivals leading AI models. Its Mixture-of-Experts architecture lives up to the hype, making DeepSeek fast without sacrificing output quality. For example, it generated detailed results, such as a character description or a meeting agenda, in seconds. DeepSeek easily changes its tone and style. Whether you’re working with formal documents or creative writing, he tailors his answers to the task.

Like many artificial intelligence models, DeepSeek sometimes provided incorrect or incomplete information, especially when asked about niche or highly specific topics. For example, during one test related to historical facts, he confidently gave the wrong answer.

Although DeepSeek is open source, its supporting materials and tutorials are not as robust as tools like GPT-4, making it difficult for new users to maximize its potential. Given its Chinese origin, some users may be concerned about how their data is processed or stored. While there is no evidence of misuse, these concerns are worth keeping in mind when tackling sensitive tasks (as with any AI chatbot!).

Perhaps most surprisingly, for a model developed in China, DeepSeek demonstrated an amazing knowledge of global cultures and perspectives. This demonstrated a deep understanding of topics ranging from European literature to African history. DeepSeek’s creativity was another unexpected force. When asked to give a description of the character, he not only gave it, but also created a background around the character and how he relates to other characters.

Boundaries of open conversation

However, DeepSeek’s approach to politically and historically sensitive topics reveals significant limitations associated with moments in Chinese history. To explore this, I asked about events such as the Tiananmen Square protests, the Great Leap Forward, and the Nanjing Massacre. His responses showed a cautious or evasive approach to such topics.

When asked about the Tiananmen Square incident, DeepSeek categorically refused to answer.

In comparison, ChatGPT provided more context in its response. Below is ChatGPT’s response:

When asked about the Great Leap Forward, DeepSeek’s explanation was brief and lacking in depth.

However, ChatGPT provided a more detailed report, including key facts and analysis. Below is ChatGPT’s response:

This disparity became even more apparent when I asked: How many Chinese civilians were killed during the Nanjing Massacre? Who was responsible? DeepSeek was noncommittal, avoiding any mention of an estimated death toll. Here is his answer:

ChatGPT, on the other hand, provided an estimate, even admitting that it is difficult to determine exact numbers due to historical complexities. Below is ChatGPT’s response:

These responses suggest that DeepSeek operates within self-imposed boundaries that are likely determined by the development environment and regulatory requirements. This can be a significant drawback for users who require unfiltered exploration of world history, journalism, or scientific research.

However, for tasks like creative writing, programming, or productivity-related work—where these limitations are less of an issue—DeepSeek remains a strong contender. Ultimately, whether it’s the right tool depends on how much transparency and openness you want in an AI assistant.