OpenAI’s new flagship crushes coding tests and complex logic, but creative work is sterile, safety rails are heavy-handed, and its cramped context window leaves it lagging rivals like Claude.

Following the rollout of OpenAI's latest GPT-5 model earlier this week, a certain user base was adamantly calling for the return of the previous GPT-4o model. Outspoken users complained about the writing quality of the updated model,