A Copyleaks research paper estimated that 74% of the outputs from rival Chinese AI model, DeepSeek-R1, were classified as written by OpenAI.

AI models are numerous and confusing to navigate, but the benchmarks used to measure their performance are also challenging.