After using Claude Opus 4.7… yes, performance drop is real.

About this article

After 4.7 was released, I gave it a try. A few things that really concern me: 1. It confidently hallucinates. My work involves writing comparison articles for different tools, so I often ask gpt and it to gather information. Today I asked it to compare the pricing structures of three tools (I’m very familiar with), and it confidently gave me incorrect pricing for one of them. I honestly don’t understand why an upgraded version would make such a basic mistake. 2. Adaptive reasoning feels more ...