Asianfin – At the 2024 Yunqi Conference, Alibaba Cloud CTO Zhou Jingren announced the upgrade of its flagship model, Qwen-Max, which now approaches the performance of GPT-4o. The Qwen-Max model has been integrated into both the Tongyi website and Tongyi APP, continuing to offer free services to all users. Additionally, users can access the Qwen-Max API via Alibaba Cloud's Bai Lian platform. Compared to its predecessor, Qwen-Max benefits from an increased volume of training data, a larger model scale, and enhanced alignment with human preferences, leading to significantly higher intelligence levels. In authoritative benchmarks such as MMLU-Pro, MATH, GSM8K, MBPP, MultiPL-E, and LiveCodeBench, Qwen-Max's performance is comparable to GPT-4o, with notable improvements in mathematical and coding capabilities that surpass GPT-4o. Since the release of the original Tongyi Qwen model in April 2023, Qwen-Max has achieved a 46% improvement in understanding, a 75% boost in mathematical abilities, a 102% increase in coding proficiency, and a 35% enhancement in resistance to hallucinations. Additionally, instruction adherence has surged by 105%, and the model’s alignment with human preferences has grown by over 700%.