Alibaba's open source large model Qwen-2-72B beat the crowd and won the top spot in the global open source large model rankings, demonstrating China's significant progress in the field of artificial intelligence. This ranking was re-evaluated by the Huggingface platform, and the evaluation criteria are more stringent and can better reflect the true performance of the model. Qwen-2's leading position is not only an affirmation of Alibaba's technical strength, but also represents the rise of China's open source large models on the international stage.
Chinese technology giant Alibaba’s latest open source fine-tuned version of the Qwen-2-72B command has topped the global open source large model rankings.
According to re-evaluation data from huggingface, a world-renowned open source platform, Qwen-2 defeated Llama-3 from Meta and Mixtral from France's Mistralai, demonstrating China's leadership in the field of open source large models.
This re-evaluation of the rankings is to provide a more objective ranking. The evaluation process has become more difficult and aims to reveal the true strength of each model. The data shows that Qwen-2 ranks first, the Llama-3-70B fine-tuned version ranks second, and the Mixtral-8x22B fine-tuned version ranks fourth. In addition, Microsoft's Phi-3-Medium-4K14B model ranked fifth, demonstrating the potential of small parameter models.
It is worth mentioning that China’s Yi-1.5-34B-Chat version ranked sixth, and Cohere’s open source Command R+104B model ranked seventh. Overall, Chinese open source models occupy four seats in the top ten, showing strong competitiveness.
The ranking results have attracted attention and discussion among people in the industry. Some experts said that China's competitiveness in the field of open source large models has long been revealed, and Qwen-2's results are even more amazing. When competing with internationally renowned closed-source large model platforms, Qwen-2 also showed impressive strength and became the only domestic company to enter the top ten of the US evaluation standards.
Highlights:
⭐ Alibaba’s Qwen-2-72B instruction fine-tuned version tops the global open source large model rankings
⭐ China’s open source models occupy four seats in the top ten, and their leading position is stable
⭐ Qwen-2 shows its competitiveness with the internationally renowned closed-source large model platform
The success of Qwen-2 not only enhances China's reputation in the global artificial intelligence field, but also indicates the huge potential for the future development of China's open source large models. In the future, we have reason to expect more Chinese forces to make breakthroughs in the field of artificial intelligence.