[ozgune]

Ack. I took the benchmark results that AI Labs themselves published for their models. So the Opus 4.6 baseline would be from the time that Anthropic released the model.