OpenAI's GPT-5 has been launched, claiming high performance in coding and logic tasks with benchmark scores of 94.6% on math tests and 74.9% on coding tasks. However, user reactions have been mixed, with criticisms highlighting its limitations in creative writing and emotional expression. Compared to Claude 4.1 Opus, its creative outputs were described as lacking depth and engagement. GPT-5 failed to handle sensitive topics and offered limited information retrieval, especially under long-context settings. While it performed well in logical reasoning, surpassing previous models, basic arithmetic tasks revealed significant flaws. Nonetheless, it excels in producing clean, functional code, making it valuable for developers despite being less effective for creative writing and casual users. Overall, GPT-5's capabilities appear strong in structured tasks, but it struggles with the creativity and nuance that other models exhibit, leading to user demand for the return of GPT-4o as a more balanced alternative.

Source 🔗