Why does anthropic change the set of benchmarks they use with every new model release?
https://www.anthropic.com/news/claude-opus-4-7
https://www.anthropic.com/news/claude-opus-4-6
1. Benchmarks saturate 2. They select the most impressive improvments