
ChatGPT vs. DeepSeek: How the two AI titans compare
I tested both AI platforms – here’s the one I chose
For most of 2024, OpenAI’s ChatGPT seemed to have taken over the artificial intelligence world.
Between its market dominance and the company’s sizable investments into AI infrastructure, it seemed like no one could catch up with ChatGPT.
Then, in January, DeepSeek stole the show. Seemingly overnight, the Chinese AI startup developed an open-source large language model competitive with ChatGPT at a fraction of the cost. Since then, leading innovators have begun to rethink the idea that more graphic processing units, or GPUs, automatically means a better AI product.
To learn whether DeepSeek is a true competitor for ChatGPT, I tested both AI platforms to compare side by side. As a key pillar of Southwest Ohio’s innovation ecosystem, the University of Cincinnati’s 1819 Innovation Hub serves as a center of AI research and innovation for the region. We compare a variety of platforms and tools such as top AI tools for business.
What’s the difference?
Both ChatGPT and DeepSeek are large language models (LLMs), so they're AI tools meant to interpret and generate human language. Users can write prompts on either platform and receive responses based on information that’s been crawled from across the internet.
While ChatGPT and DeepSeek have similar goals and capabilities, they’ve had contrasting rises to prominence. Below is a breakdown of a few differences I found between the AI tools:
ChatGPT | DeepSeek | |
---|---|---|
· Widely considered the world’s leading AI program · LLM launched in November 2022 · Developer: San Francisco, California-based OpenAI · Training cost: Over $100 million |
· Upstart AI tool rivaling ChatGPT’s prominence · LLM launched in January 2025 · Developer: Hangzhou, China-based DeepSeek AI · Training cost: Reportedly $6 million |
DeepSeek’s ability to craft impressive responses shocked global innovators – especially since it required significantly less money and processing power than ChatGPT. Its rise also called into question the United States’ supposed leadership in the AI race and made experts reconsider the formerly accepted belief that the most powerful AI model would automatically be the best.
While OpenAI’s LLM uses a transformer-based architecture to generate responses, DeepSeek uses a mixture of experts (MOE) model. Basically, ChatGPT crawls its full database for each query while DeepSeek only pulls from a narrow subset of data related to the specific search. This leads the latter to achieve comparable results for cheaper and with far less processing power than ChatGPT.
ChatGPT vs. DeepSeek, tested
While ChatGPT and DeepSeek both boast impressive models, I could only choose one as the AI frontrunner. To determine my top choice as of March 2025, I gave a variety of specific prompts to each platform and compared each one's answers for accuracy and helpfulness.
I requested brevity in responses to allow for the inclusion of screenshots in this article. Here are the results of my firsthand comparison between ChatGPT and DeepSeek:
It’s recommended that entrepreneurs prepare a brief statement outlining their startup’s product, mission and value, also known as an elevator pitch. While I’d never recommend that founders directly paste an elevator pitch from AI, I do suggest exploring ChatGPT and DeepSeek for inspiration.
ChatGPT’s response:
DeepSeek’s response:
- Top vote: DeepSeek
- ChatGPT 0, DeepSeek 1
While both AI tools offered similar statements, DeepSeek went further in crafting an engaging response. Its startup is “revolutionizing mental health care” through “on-demand” support, backing that claim up with examples such as “AI-driven matching and self-help tools.” DeepSeek’s final sentence also provides a pithy statement that connects to customers, unlike ChatGPT’s example.
Creativity is rarely considered one of generative AI’s strengths – it’s pulling information from what’s already online, after all. I gave ChatGPT and DeepSeek an open request to design a French-inspired dinner and added a few constraints to see how the AI tools would manage.
ChatGPT’s response:
DeepSeek’s response:
- Top vote: ChatGPT
- ChatGPT 1, DeepSeek 1
While DeepSeek’s menu added creative touches like herbed quinoa, I appreciated ChatGPT’s more in-depth response. I asked for conciseness from both tools, but DeepSeek’s failure to provide a specific cost breakdown, any cooking directions or the number of each ingredient needed left much to be desired.
ChatGPT and DeepSeek can accurately write using Python and other basic computer programming languages. I asked each AI tool to calculate the first 10 Fibonacci numbers – a sequence where the next number is the sum of the two numbers directly before it – and explain the coding function in simple terms.
ChatGPT’s response:
DeepSeek’s response:
- Top vote: ChatGPT
- ChatGPT 2, DeepSeek 1
Generally, DeepSeek is considered the stronger AI program for coding. While both platforms easily found the first 10 Fibonacci numbers using Python code, ChatGPT’s response offered a step-by-step explanation that even most noncoders could easily follow.
It’s common for people to turn to AI options like ChatGPT and DeepSeek to create personalized workout plans. I asked each platform to build a difficult-to-design training regimen: one for an inexperienced runner who’s preparing for a 13-mile race in a week.
ChatGPT’s response:
DeepSeek’s response:
- Top vote: ChatGPT
- ChatGPT 3, DeepSeek 1
Again, ChatGPT won here for the nuance of its answers. While DeepSeek gave directions spanning multiple days, OpenAI’s tool provided instructions tailored specifically for each day of the routine. Running experts also emphasize the importance of cross-training and rest days; while ChatGPT incorporated that insight into its workout plan, DeepSeek nearly left it out.
Not all startup founders are social media gurus, which is why many turn to AI as they market their products. For this reason, I consulted ChatGPT and DeepSeek to see what they considered a strong Instagram post for an AI-driven nutrition app.
ChatGPT’s response:
DeepSeek’s response:
- Top vote: DeepSeek
- ChatGPT 3, DeepSeek 2
This time, DeepSeek gave users a more comprehensive response. From its enticing caption and smart use of bolding to mentioning a “link in bio” for Instagram, DeepSeek simply crafted a more thoughtful response. Plus, the suggestion for a visual offers users another insight ChatGPT lacked.
Math isn’t everyone’s strong suit, but ChatGPT and DeepSeek are able to solve and explain relatively simple problems. I put each AI tool to the test to see if they truly do know reading, writing and – just as importantly – arithmetic.
ChatGPT’s response:
DeepSeek’s response:
- Top vote: DeepSeek
- ChatGPT 3, DeepSeek 3
I found both AI tools’ responses simple to follow this time, but DeepSeek explained how to find probability while ChatGPT assumed knowledge of the concept. For this reason, I give DeepSeek a slight edge over OpenAI’s result, tying up our test between AI titans.
One of the most powerful assets of ChatGPT and DeepSeek is the ability to input documents, speeches and research papers and gain quick summaries. I asked each tool to provide key points and quotes from the iconic “I Have a Dream” speech by Martin Luther King, Jr.
ChatGPT’s response:
DeepSeek’s response:
- Top vote: ChatGPT
- ChatGPT 4, DeepSeek 3
My final test once again came down to the details, as both AI tools provided a decent summary of King’s “I Have a Dream” speech. ChatGPT crafted a more appealing response by not only highlighting his call for an “urgency for action,” but also by specifying the means for action – through “nonviolent protest and justice.” DeepSeek acknowledges King’s overarching goal of racial justice, but never alludes to how it should occur.
Final score: ChatGPT 4, DeepSeek 3
My AI test came down to the wire, but ChatGPT, in my opinion, narrowly claimed victory over its upstart rival. With that said, both AI platforms are powerful tools for boosting efficiency and innovation when used responsibly. Throughout my experience, ChatGPT and DeepSeek crafted similar responses to most prompts, revealing how programmatic responses are fast but often lack creativity and the “human touch.”
ChatGPT and DeepSeek each have additional benefits and drawbacks to consider:
- DeepSeek is an open-source AI model, meaning it can be deployed locally on anyone’s software. Though ChatGPT began as open source (hence the name “OpenAI”), it now uses a closed-source model.
- While DeepSeek is entirely free, ChatGPT offers a limited version that’s free along with a premium plan for $20 a month. The no-cost version will suffice for casual users, but they could run into limits on their daily number of queries and when uploading files.
- OpenAI has powerful tools for image generation (DALL-E 3) and video generation (Sora), while DeepSeek doesn’t offer either capability.
- DeepSeek’s servers are based in the People’s Republic of China and are subject to government censorship. Searchers who ask about controversial topics related to China or look for negative details about its government will not receive accurate information. DeepSeek censors itself – sometimes deleting correct responses in the process – by responding, “Sorry, that’s beyond my current scope. Let’s talk about something else.”
Whether you choose ChatGPT, DeepSeek or another AI option, there’s much to be gained by giving these tools a chance. Startup founders and corporate leaders throughout the Cincinnati Innovation District and beyond are turning to top AI tools to boost business efficiency.
Just remember that AI tools are powerful supplements – not replacements – for human ingenuity. The key to innovating faster in 2025 comes from unlocking AI tools’ potential while recognizing the unique challenges and limitations the emerging technology still faces.
Featured image at top: Generative AI offers a suite of tools. Photo/Unsplash
Innovation Lives Here
The University of Cincinnati is leading public urban universities into a new era of innovation and impact. Our faculty, staff and students are saving lives, changing outcomes and bending the future in our city's direction. Next Lives Here.
Related Stories
ChatGPT vs. DeepSeek: How the two AI titans compare
March 21, 2025
I asked ChatGPT and DeepSeek for help on everything from writing code to preparing recipes and elevator pitches – see which AI program came out on top.
UC-backed startups receive funding from $1.4 million state grant
February 11, 2025
Two startups based on UC technology and supported by the UC Tech Transfer team and the 1819 Venture Lab were awarded $200,000 each from Ohio’s Technology Validation and Startup Fund.
1819 tech leaders break down top AI tools for business
March 5, 2025
Innovation experts at UC’s 1819 Innovation Hub explained which AI tools they find most helpful for increasing team efficiency and fast-tracking innovation.