Opus 5.5 is the smartest model I've used but it fucking forgets everything 😭

A user on reddit.com praises Opus 5.5 for its impressive reasoning and ability to provide working solutions, often on the first attempt. However, they express frustration that each session starts from scratch, requiring them to re-explain their project's stack, structure, and conventions. They also note that while the model performs well for 30 to 40 messages mid-session, it then begins to drift and forget information.

This user review highlights Opus 5.5's advanced reasoning compared to other models, but uniquely details its session memory limitations after 30-40 messages.