making history, alibaba tongyi’s open source model qwen2.5 entered the top ten in the world in the blind test of large models
2024-09-30
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
chao news client reporter zhang yunshan
according to news on september 29, the benchmark testing platform chatbot arena recently announced the latest blind test list of large models. the alibaba tongyi qianwen open source model qwen2.5 released 10 days ago once again broke into the top ten in the world. its large language model qwen2.5 -72b-instruct ranks tenth on the llm list and is the only chinese large model in the top ten; qwen series visual language model qwen2-vl-72b-instruct ranks ninth on the vision list and is the highest-scoring open source large model.
at the same time, the number of derivative models developed by the global open source community based on the secondary development of the qwen series exceeded 74,300, surpassing the 72,800 derivative models of the llama series. tongyi qianwen qwen has grown into the world's largest generative language model family. on the open llm leaderboard, the authoritative list of open source models in the hugging face community, the qwen series and its derivative models have occupied all the top ten seats.
whether it is model performance or ecological influence, qwen has created the history of open source large models in china.
qwen2.5-72b-instruct ranks tenth on the chatbot arena large language model list
chatbot arena is a large model performance testing platform launched by the open research organization lmsys org. since its launch in may 2023, it has been the most important arena for the world's top large models. the platform currently integrates more than 70 large models around the world. the large models are anonymously divided into pairs and handed over to users for blind testing. users vote on the model capabilities based on real conversation experience.
qwen2.5, released on september 19, quickly entered the list. the score of the flagship model qwen2.5-72b-instruct ranked tenth on the llm list, behind openai's o1, gpt-4o and other models, and is the chinese large model with the highest score. ; qwen2-vl-72b-instruct, an open-source visual language model on the same day, broke into the ninth place on the vision list, slightly behind closed-source models such as gpt-4o and gemini-1.5-pro, and is the best-performing open-source model. previously, several open source models in the qwen series have entered the chatbot arena list.
chatbot arena officially announced that qwen2-vl-72b-instruct is the highest-ranking open source visual language model
the release of qwen2.5 triggered a carnival in the open source community at home and abroad. this set of open source models covers large language models, multi-modal models, mathematical models and code models of multiple sizes. almost all sizes of models have achieved the same scale in the industry. best performance, more than 1.5 million downloads in 10 days of release. some foreign developers praised tongyi qianwen as a true "open ai"; some user reviews found that qwen2.5 is sota level from 0.5b to 72b, so they started a topic: "everyone has replaced it with qwen2.5 what are the commonly used models?”
overseas open source communities praise qwen2.5
overseas open source communities praise qwen2.5
"please indicate the source when reprinting"