On May 6, the reporter learned that LiveBench, an authoritative international model evaluation list, announced the latest ranking. Alibaba's open source next-generation Tongyi Qianzhen model Qwen3 won the global open source model championship, and surpassed top closed source models such as O3 High, O4-mini High, and Gemini 2.5 pro in following this key ability, ranking first in the world. According to information, the LiveBench list was launched by Turing Award winner Yang Likun, Meta's chief AI scientist, in collaboration with New York University and other institutions to comprehensively evaluate large models from multiple complex dimensions such as mathematics, reasoning, programming, and language understanding. Because it uses a dynamically updated question bank, it is known as “the world's first uncheatable model benchmark.”

On May 6, the reporter learned that LiveBench, an authoritative international model evaluation list, announced the latest ranking. Alibaba's open source next-generation Tongyi Qianzhen model Qwen3 won the global open source model championship, and surpassed top closed source models such as O3 High, O4-mini High, and Gemini 2.5 pro in following this key ability, ranking first in the world. According to information, the LiveBench list was launched by Turing Award winner Yang Likun, Meta's chief AI scientist, in collaboration with New York University and other institutions to comprehensively evaluate large models from multiple complex dimensions such as mathematics, reasoning, programming, and language understanding. Because it uses a dynamically updated question bank, it is known as “the world's first uncheatable model benchmark.”