We tested Gemini 2.5 Pro with refact.ai on Polyglot Benchmark The model from Google got  82.2%, behind Claude 3.7 Sonnet’s 93.3% In the video I am providing the details how our model tests are different from aiders