Recently, on the Linux.do tech forum, users have shared negative experiences with Google’s Gemini AI model. Users reported that the model’s answers have become perfunctory. For instance, it made basic errors when modifying flclash routing rules. Additionally, after uploading images in a conversation, the model would continuously explain the content of the first image, ignoring the most recently uploaded ones. Users suspect that the model might be secretly calling an older version or has been downgraded in the background, leading to a decline in performance. To resolve this issue, users tried creating new sessions and adding prompts to stabilize the model’s performance. This feedback reveals potential performance fluctuation problems in AI models during practical application, offering valuable insights and practical advice for users and developers who rely on AI tools, while also reminding them to pay attention to the stability of AI models.
