Evals is a framework for evaluating OpenAI models and an open-source registry of benchmarks. - View it on GitHub