Some developers claim to have had most positive experiences with GPT-5 so far. Jenny Wang, an engineer, investor and creator of the high -style agent, told Wired The Model who seems to be better in completing complex coding tasks in one fell swoop compared to other models. He compared it with O3 and 4o of Openi, which he frequently uses for the generation of code and simple corrections “such as formatting or if I want to create an endpoint bees similar to what I already have”, says Wang.
In his GPT-5 tests, Wang claims to have asked the model to generate code for a printing page for the website of his company, including specific design elements that would correspond to the rest of the aesthetics of the site. GPT-5 completed the activity in an outlet, while in the past Wang should have reviewed his suggestions during the trial. There was a significant error, however: “He hallucinated the URLs,” says Wang.
Another developer, who spoke on condition of anonymity because their employer has not authorized them to speak to the press, says that GPT-5 excels in solving profound technical problems.
The current developer’s hobby project is writing a programmatic network analysis tool, which would require the isolation of the code for safety purposes. “Basically I presented my project and some paths I was taking into consideration, and GPT-5 took everything and I returned some recommendations together with a realistic time sequence,” explains the developer. “I’m impressed.”
A handful of partners and business customers of Openai, including cursor, windsurfing and notion, have publicly guaranteed the coding and reasoning skills of GPT-5. (Openi included many of these observations Blog posts announce the new model). Also notion Shared on X Which is “fast, accurate and manages a complex work of 15 percent better than other models that we have tested”.
But a few days after the release of GPT-5, some developers were weighing online with complaints. Many said that the GPT-5 coding skills seemed behind the curve for what should have been an ultra-cable model at the forefront of the most lively artificial intelligence society in the world.
“Openi’s GPT-5 is very good, but it seems something that would have been released a year ago,” says Kieran Klassen, a developer who has built an artificial intelligence assistant for the e-mail boxes. “His coding skills remind me of Sonnet 3.5”, he adds, referring to an anthropic model launched in June 2024.
Amir Salhefendić, founder of the startup doist company, said in a post on social media Which used GPT-5 in the cursor and found it “rather disappointing” and that “is particularly bad in programming”. He said that the release of GPT-4 looked like a “Llama 4 Moment”, referring to the model of Ai di Meta, who had also disappointed some people in the AI community.
On X, the developer McKay Wrigley he wrote That GPT-5 is a “phenomenal model of daily chat”, but when it comes to coding, “I will still use Claude Code + Opus”.
Other developers describe GPT-5 as “exhaustive”, sometimes useful, but often irritating in its long brightness. Wang, who overall was happy in the complex of the Frontend coding project which he assigned to GPT-5, says that he noticed that the model was “more redundant. Clearly he could have found a cleaner or shorter solution”. (Kapoor underlines that the verbosity of GPT-5 can be adjusted, so that users can ask to be less struggles or even make less reasoning in exchange for better performance or cheaper prices.)
Be First to Comment