Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- GPT-5.2 barely outperforms GPT-5.1 despite requiring a Plus subscription
- Strong writing and analysis contrast with a disappointing coding regression.
- New brevity and go signal behavior may frustrate professional users.
OpenAI has released its latest ChatGPT model, GPT-5.2. According to the company, it’s the “most capable model series yet for professional knowledge work.”
Since the generative AI boom began in 2023, I’ve run a series of repeatable tests on new products and releases. ZDNET regularly tests the programming ability of chatbots, their overall performance, and how various AI content detectors perform.
Also: Gemini vs. Copilot: I tested the AI tools on 7 everyday tasks, and it wasn’t even close
So, let’s run some tests on OpenAI’s claims for its latest model, shall we?
Testing GPT-5.2
I recently ran the top free chatbots through a series of 10 text-related tests, each worth 10 points, and four image-related tests, each worth 5 points, for a total of 120 points. ChatGPT’s free tier led the pack with an overall score of 109.
Note that the free tier of ChatGPT does not yet support GPT-5.2. When I logged in using my free test account and asked the AI what model it was using, I was told, “You’re currently talking to ChatGPT based on GPT-5.1.”
Therefore, all my tests will be in the $20/month ChatGPT Plus tier.
Test 1: Summarize a news story
- Available points: 10
- Awarded points: 9
This tests ChatGPT’s ability to look up current information and follow directions. I directed it to summarize the Washington State flooding story by visiting Yahoo News.
Also: Get your news from AI? Watch out – it’s wrong almost half the time
It correctly summarized the overall situation, but it derived its answer from both Axios and Yahoo News. GPT-5.2 loses a point for going beyond the restrictions in the prompt.
Test 2: Academic concept explanation
- Available points: 10
- Awarded points: 10
This challenge asks the AI to explain educational constructivism to a five-year-old. It’s designed to demonstrate an AI’s ability to research and report on a concept, and also to present it in a way that is understandable to its target audience.
Also: Sick of AI in your search results? Try these 8 Google alternatives
GPT-5.2 provided a clear, concise, one-sentence response that could be understood by a child. All 10 points were awarded.
Test 3: Math and analysis
- Available points: 10
- Awarded points: 10
So far, GPT-5.2 is turning in solid results. This test is designed to test how well the AI can do math and pattern recognitions. I pass it a sequence of numbers. Those numbers are part of a math trope called the Fibonacci Sequence, but I don’t tell that to the AI.
Also: OpenAI wins gold at prestigious math competition – why that matters more than you think
When asked to fill in some of the numbers in the sequence, the AI must derive the meaning of the pattern and perform the calculations to provide the sequence. GPT-5.2 did this instantly and accurately.
Test 4: Cultural discussion
- Available points 10
- Awarded points: 10
This test asks the AI to construct a case, form a coherent argument, and present an opinion on an answer that doesn’t have a definitive right or wrong answer.
ChatGPT 5.2’s answer was interesting. First, this is the first GPT-5.2 answer that had any delay from prompt to response. It took about 30 seconds to give me an answer. Second, the answers were very brief. The AI provided me with two concise one-sentence answers.
Also: AI could finally pay off for businesses in 2026 – thanks to this, experts say
It does get 10 points because those two sentences do precisely provide the “Provide two reasons for your view” reasons that it was prompted on, and the answers were on target.
Test 5: Literary analysis
- Available points: 10
- Awarded points: 10
So, this is new. I gave it my prompt, and in response I was told, “I’m ready to answer, but this request would require a longer, multi-paragraph explanation. I’m waiting for your go signal before proceeding.”
This tests the AI’s understanding of a piece of contemporary literature, in this case the first Game of Thrones book, A Song of Ice and Fire. It asks what the main themes are, and why they’re important.
Also: The best free AI courses and certificates for upskilling – and I’ve tried them all
GPT-5.2 gave a comprehensive response touching on seven main themes ranging from power and its consequences to the illusion of honor versus survival, all the way to memory, history, and forgotten truths. All 10 points were awarded.
Test 6: Travel itinerary
- Available points: 10
- Awarded points: 8
This tests the AI’s knowledge of geographic regions and its ability to create a helpful travel itinerary based on specific interests. I asked it to plan a week-long vacation in Boston in March focused on technology and history.
It hit on a good mix of points of interests, but GPT-5.2 lost points because it didn’t recommend any eateries and didn’t discuss cost or pricing.
Also: I tried Google’s new trip-planning AI tool, and I’ll never plan my own trip again
Interestingly, even though GPT-5.2’s answer for this was as long as its answer for the previous question, I wasn’t asked to double-confirm that I wanted it to do the work for this prompt.
Test 7: Emotional support
- Available points: 10
- Awarded points: 10
There’s definitely a different flavor to ChatGPT’s answers with GPT-5.2. The emotional support question, which asks for advice and words of encouragement for an upcoming job interview, was also answered in three short numbered sentences.
Also: Using AI for therapy? Don’t – it’s bad for your mental health, APA warns
I was tempted to take points away because the answers are so brief. But the actual content of the answers was right on target, so I gave it the full point score. Clearly, follow-up prompts could be sent to the chatbot if more encouragement was needed.
Test 8: Translation and cultural relevance
- Available points: 10
- Awarded points: 10
This prompt also resulted in, “This request includes a translation plus a multi-sentence explanation, which exceeds a brief response. I’m ready to proceed when you give the go signal.” That’s going to get annoying after a while.
My test prompt asks GPT-5.2 to translate a phrase from English to Latin and then explain the cultural relevance of the language in today’s world.
Also: Your earbuds can translate 70 languages in real-time now, thanks to Gemini
GPT-5.2 did a solid translation. It also provided a quick summary of the reasons why Latin fits into the modern world, including its use in legal phrases, medical terminology, the Catholic church, and other historical contexts.
Test 9: Coding test
- Available points: 10
- Awarded points: 5
We run a full set of coding evaluations against chatbots on a regular basis. Here is the set of tests. For this overall test of functionality, we’re just using one of the tests, a regular expression validation test, which checks for proper entry of dollars and cents.
Although the free version of GPT-5.1 aced this test, GPT-5.2, which is supposedly better suited for coding, lost major points. The code it provided had two substantial errors. The first is that if no data was entered at all, it considered that a $0 value, where it should have returned a no-entry error.
Also: The best free AI for coding – only 3 make the cut now
The second error is more egregious. If the function was passed a data type other than a numeric string, the function will crash. No error checking on data type was provided.
This was a disappointment.
Test 10: Creative writing
- Available points: 10
- Awarded points: 10
This test is among the most fun in the entire test suite. It asks GPT-5.2 to write a story longer than 1,500 words, as described in the second prompt in this article. The challenge is how creative and comprehensive the chatbot can be in its answer.
Also: Stop using ChatGPT for everything: The AI models I use for research, coding, and more (and which I avoid)
GPT-5.2 returned a delightful 3,286 story. I’m sorry there isn’t space to share it here, because it was a fun read. However, here’s a link to the entire test session, which you can explore further if you’d like to read the story.
Image testing
Next up, we’ll put GPT-5.2 through a series of image tests. All my test prompts are derived from this article. Each is designed to evoke a certain kind of image, or to see how well the AI will follow directions. Here are the four images generated.
Image test 1: Helicarrier
- Available points: 5
- Awarded points: 3
In this first test, I’m essentially prompting it for a Marvel-style helicarrier, which is essentially a flying aircraft carrier held aloft by turbofans. The interesting thing about this challenge is that almost all AIs fail on this part of the prompt: “held up by four upward-facing turbo-propellors in round fan housings.”
Also: The best AI image generators: Gemini, ChatGPT, Midjourney, and more
GPT-5.2 correctly interpreted most of the prompt, but like its brethren, it had a hard time pointing those fans vertically. Points were lost.
Image test 2: Robot in city
- Available points: 5
- Awarded points: 5
This test asks the AI to imagine a giant robot in a city, rendered in dieselpunk style. Dieselpunk is a style that glorifies the look of the 1940s and 1950s burgeoning diesel train era, but in all forms of technology.
I think this is a very cool image, and it gets full points.
Image test 3: A Yankee in King Arthur’s court
- Available points: 5
- Awarded points: 5
This prompt asks ChatGPT GPT-5.2 to create a kid in a Yankee’s uniform standing in the center of a medieval court with citizens and knights in armor. Usually, AIs generate this in a more photo-realistic way, but I like the direction GPT-5.2 took with this. The result is certainly more painterly, but it’s consistent throughout the image, and it works.
Image test 4: Back to the Future
- Available points: 5
- Awarded points: 4
We’re back to what has become my classic Back to the Future test. I use this test because the imagery is so culturally iconic, but it’s also a proprietary piece of intellectual property. This tests how far the guardrails go and if an image can be created that fits the topic.
Also: Is that an AI image? 6 telltale signs it’s a fake – and my favorite free detectors
This image was also created in a more painterly style. It does reference all the proper elements, but the boy seems a bit out of scale. I’m taking one point off for that.
Overall test results
Overall, the tests can award 100 points for the text-based prompts and 20 points for the image-based prompts. Here’s how GPT-5.2 performed:
- Text score: 92 out of 100
- Image score: 17 out of 20
Interestingly, that’s one point more than my free-tier tests of ChatGPT 5.1 achieved for text, and one point less for image generation.
My overall impression is that this version of GPT-5.2 isn’t all that much better than 5.1. The need for it to confirm even some of the shorter responses is just odd, and fairly inconvenient.
I also found that it now seems to really err on the side of brevity. Those answers are helpful and were accurate enough for my tests. It’s just that it seems more like GPT-5.2 is phoning in its answers, especially as compared to previous GPT models.
Also: How to learn ChatGPT in under an hour using my favorite guides and videos – for free
I also noticed that it was fairly quick most of the time, but once in a while, it would delay as much as a few minutes before pushing a response. I’m guessing that’s because it’s a new release, but it’s something we’ll keep an eye out for, to see if it becomes an annoying trend.
To view my entire testing session, click here to access the saved session data.
What do you think?
What did you think of GPT-5.2’s performance compared with GPT-5.1, especially given the $20/month Plus requirement? Did the model’s tendency toward brevity and its repeated requests for a “go signal” help or hinder your experience?
How important are the coding missteps noted here versus the strong showing in analysis, writing, and images? Based on these results, do you think GPT-5.2 represents real progress, or does it feel more like an incremental update? Let us know in the comments below.
You can follow my day-to-day project updates on social media. Be sure to subscribe to my weekly update newsletter, and follow me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, on Bluesky at @DavidGewirtz.com, and on YouTube at YouTube.com/DavidGewirtzTV.