Alibaba’s Qwen2-72B-Instruct ranked only behind GPT-4o and Claude 3.5 Sonnet – both closed-source – based on metrics such as coding and text comprehension.