The article compares GPT-4o to Sonnet from Anthropic. I'm wondering how Opus would perform at this test?