Indeed, they probably would never admit it, but it's common-sense because why not? It's not even illegal, it makes better models, better competitor analysis... of course Anthropic is fully aware of gpt-5.5 capacity, they must benchmark it 24/7.