The Stanford Institute for Human-Centered Artificial Intelligence (HAI), led by renowned AI scientist Fei-Fei Li, has released its latest report, the 2025 AI Index Report. This authoritative report, published annually for eight years, highlights that the performance gap between top AI large language models (LLMs) from China and the United States – the two most influential AI nations globally – has significantly narrowed to just 0.3%. This is a dramatic decrease from 17.5% in 2023, rendering the difference almost negligible.
The report identifies the Notable Models of 2024. Among the 61 selected models, Google and OpenAI each have 7 models, tying for first place. Alibaba follows closely with 6 models, securing the third global position in model contribution and holding the title of the Chinese tech company with the most models included. Alibaba's six selected models are: Qwen-72B, Qwen1.5-72B, Qwen2-72B, Qwen2.5-72B, Qwen2.5-32B, and QwQ-32B.
Furthermore, Alibaba's Qwen2, Qwen2.5, and DeepSeek-V3 were selected among the 32 most representative models, datasets, and feature releases of 2024, as highlighted by the Stanford report.
The report also emphasizes Alibaba's contributions to open-source AI. Since 2023, Alibaba's DAMO Academy has open-sourced over 200 models. Currently, the number of models derived from Qwen has exceeded 100,000, surpassing Llama from the US to become the world's largest open-source large model ecosystem.
2025 AI Index Report website:https://hai.stanford.edu/ai-index/2025-ai-index-report