Every time a new AI model is released, it’s typically touted as acing its performance against a series of benchmarks. OpenAI’s GPT-4o, for example, was launched in May with a ...