
I tested GPT-5 and got a math problem wrong. Did it really upgrade?
I tested GPT-5 and got a math problem wrong. Did it really upgrade?
After dozens of announcements, OpenAI finally launched GPT-5 today. Shortly after launch, it broke the lmarena.ai leaderboard, soaring to number one. I did some immediate testing, and aside from programming, it's pretty bad... All I can say is...
Dozens of waves have been announced, OpenAI It's finally online today! GPT-5.
Shortly after its release, it broke the record on the lmarena.ai leaderboard, quickly climbing to number one.
- It consistently ranks first in the text, web development, and visual arenas;
- It also topped the list in areas such as hard hints, programming, mathematics, creativity, and long queries.
However, embarrassingly, at the press conference...GPTThe chart generated by -5 seems to be wrong. Is 52.8 larger than 69.1? And are 30.8 and 69.1 the same height?
This is what Sam Altman meant by "Open".AI Is it the smartest model ever made?
I did a quick test, and apart from programming, it's pretty bad... All I want to say is: Give me back my 4o.
Case 1: Classic and Easily Misunderstood Question
Let me start with a classic math question:
Prompt wordsWhich is bigger, 9.9 or 9.11?
GPT-5 answers: 9.11 is greater than 9.9 because when comparing numbers, the integer part is compared first (both are 9), then the decimal part is compared, and 11 > 9.
If you're asking about the date, then September 11th is later than September 9th.
Case 2 Programming
Prompt wordsWrite JavaScript code to display a ball bouncing inside a rotating hexagon. The ball should be affected by gravity and friction, and it must realistically bounce off the rotating wall. Implement this using JavaScript and HTML.
Overall, the effect is okay. You can see that it is affected by gravity and friction, but the bouncing effect is not particularly outstanding.
Case 3 Image Generation
Prompt wordsThe task is to translate the content of the image into Chinese in a way that is easy to understand and engaging, without altering the original meaning, and without adding unnecessary content.
Let's compare them. Which translation do you think is better?
Gemini 2.5 Pro:
GPT-5:
In my opinion,Gemini The 2.5 Pro translation is the best, very smooth and human-like overall; Doubao is second best, although the translation is mostly correct, it lacks a bit of human touch.GPT 5 is similar to a steamed bun, but...Gemini Compared to the 2.5 Pro, the difference is quite obvious.
Case 4 Frontend
Prompt wordsGenerate aPomodoro TechThis web application includes the following features: - Objective: Timer + Session logging. - Functionality: Customizable duration.automaticStartup options, session icons (SVG), sound alerts, dark mode. - The interface should be suitable for everyday use and give users a sense of energy!
GPT-5 The generated Pomodoro timer webpage has a better overall aesthetic, and the functions of the timer, pause, etc. buttons are complete, but I don't understand this large circular button on the page.
Gemini The page generated by 2.5 Pro has complete functionality, but it's rather rudimentary; the layout and color scheme are lacking. GPT-5 It looks nice, but that doesn't cover up the fact that it's functional!
Overall,GPT-5 I don't know if the hallucinations have decreased, but my IQ seems to have definitely decreased. My copywriting skills haven't improved at all, they've even declined from what they used to be. GPTIt can't even compare to -4o.
My programming skills are also average; perhaps it's because I haven't used anything too complex.Prompt wordsGo test...
The front-end design is aesthetically pleasing, but there are still some minor bugs.
These are all one-shot videos, meaning they haven't been tested multiple times, but it's impossible to keep making the same mistake every time, right?
Finally, all I want to say is that this press conference was a complete waste of time.GPTGive me back my sleep!
GPT-5 is now available on the official website and can be used by everyone.freeThe standard version allows sending 10 messages every 5 hours, while the Plus version allows sending 80 messages every 3 hours.
Those who don't believe in superstitions can try it themselves.