Aya Expanse has been recognized as the best open-weights model on Scale AI's private multilingual protocol. The model reportedly outperforms proprietary models, larger models, and those developed by more researchers with greater infrastructure in certain languages. Additionally, private evaluations indicate that Aya Expanse demonstrates strong multilingual performance despite being a smaller model. There are ongoing discussions in the AI community regarding the effectiveness of current benchmarking methods, with some experts suggesting that traditional benchmarks may not accurately reflect the capabilities of modern models. The need for new benchmarks that account for randomized or generated models is also emphasized, as current benchmarks may not fully capture the practical performance of these systems.