This Formula 1 season, I'm putting frontier models to the test - can they use their research capabilities to accurately predict outcomes of races?