I tested GPT-5.2 and the AI model's mixed results raise tough questions
SMRTR summary
OpenAI's new GPT-5.2 model, requiring a $20 monthly subscription, delivered mixed results in comprehensive testing that scored it just one point higher than the free GPT-5.1 version. While excelling at writing and analysis tasks, GPT-5.2 showed concerning coding regressions and introduced annoying behaviors like requesting "go signals" for longer responses and providing overly brief answers.
SMRTR provides this summary for quick context. The original article belongs to ZDNet.
Read the original article