Qwen2-72B

Qwen2-72B tops open-model leaderboards for multilingual tasks, achieving 84.2% on MMLU while supporting 27 languages. Its strength in Chinese, Arabic, and Southeast Asian languages is 18% higher than the average Western model's performance. Released under Apache 2.0, it outperforms #9 Command R+ across most general benchmarks, but lacks specialized enterprise citation tools.