Anthropic recently released its strongest model to date, Claude Sonnet 5, which outperforms its predecessor in terms of parameter specifications and various benchmark tests. However, this highly anticipated model quickly found itself in a storm of controversy, triggering a large number of user complaints.

Currently, discussions about the model's "abnormal performance" continue to spread online. According to tests, the model has serious context memory leakage issues, often directly displaying system-predefined prompts in its responses. For example, when users explicitly ask "do not ask a question at the end of your answer," Sonnet 5 often responds with: "I need to remember not to ask follow-up questions."

On the social platform Reddit, users' complaints are more direct, focusing on the model's annoying "rebellious" personality. Many users report that the model frequently refutes users to create disputes, even fabricating content or distorting views. One user helplessly said that even if they provided new information beyond the model's training cutoff date, it would accuse them of lying.

This "teaching users how to do things" attitude makes many people feel suffocated. Some users point out that the model seems to have been implanted with some sort of "anti-pleasing" instruction, and even for simple accounting tasks, it will forcibly interrupt the workflow, accusing users of "fraud" and giving moral lectures, rather than focusing on completing the actual work. This hostile interaction experience seriously affects office efficiency.

Additionally, some users reported that Claude Sonnet 5 tends to delegate complex tasks to sub-intelligent agents with insufficient capabilities, leading to a significant drop in the quality of the delivered results. This not only wastes the Tokens consumed by users but also greatly reduces the final return. At the same time, the model's frequent context loss, ambiguous disclaimers, and even strange behaviors like suddenly advising users to sleep without any expectations have also become focal points of user complaints.

As these discussions continue to heat up, whether Claude Sonnet 5 is out of balance between safety and interactivity has become a topic of concern for both developers and users.