Some developers say they have had positive experience with GPT-5 so far. Jenny Wang, an engineer, investor and creator of personal styling broker Alta, told Wired Alta that the model seems to be better at completing complex coding tasks than other models. She compared it to OpenAI’s O3 and 4O, which she often uses for code generation and direct fixes “like formatting, or if I want to create similar API endpoints to the ones I already have,” Wang said.
Wang said in a test of GPT-5 that she asked the model to generate code by page for the company’s website, which includes specific design elements that match the rest of the website’s aesthetics. On the one hand, GPT-5 completed the task, and in the past, Wang had to modify her prompts in the process. There is a big mistake, though: “It hallucinated the URL,” Wang said.
GPT-5 is good at solving deep technical problems, and his employer does not authorize them to talk to the media, and his employer does not authorize them to talk to the media, which is another developer.
The current hobby project for developers is to write a programmatic network analysis tool that requires code isolation for security purposes. “I basically introduced my project and some of the avenues I’m considering, and GPT-5 put it all into practice and gave back some suggestions and a realistic timeline,” the developer explained. “I was impressed.”
A handful of Openai’s corporate partners and customers (including cursors, windsurfing and concepts) have publicly guaranteed the coding and reasoning skills of the GPT-5. (Openai contains many of these words in his blog post announcing the new model). The concept also shares the concept on X: “15% faster than the other models we’ve tested, which is “fast, thorough and complex work.”
But within a few days after the release of GPT-5, some developers weighed the complaints online. Many say that the encoding capabilities of GPT-5 seem to be tortuous, as they should have been the most advanced super-capacity model of the world’s most buzzing AI company.
“OpenAI’s GPT-5 is very good, but it seems like something will be released a year ago,” said developer Kieran Klassen, who has been building an AI assistant for his email inbox. He added: “Its encoding feature reminds me of SONNet 3.5.”
Amir Salihefendić, the founder of the startup Doist, said in a social media post that he has been using GPT-5 in his cursors and found it “very overwhelmed” and “especially bad when it comes to coding.” He said the release of GPT-4 felt like a “Camel 4 Moment”, referring to Meta’s AI model, which also disappointed some people in the AI community.
On X, developer McKay Wrigley wrote that GPT-5 is a “daily chat model for phenomena”, but in terms of coding, “I will still use Claude Code + Opus.”
Other developers describe GPT-5 as “exhaustive”, which can sometimes be helpful, but often annoyed by its long mood. Wang, overall, was happy with the front-end coding project assigned to GPT-5, and she said she did notice that the model was “more redundant. Clearly a cleaner or shorter solution could be proposed.” (Kapoor noted that the GPT-5 details could be adjusted so that users could ask for lower chat rates and could even take less reasoning in exchange for better performance or a cheaper price.)
