Xin Lang Cai Jing·2025-12-11 12:59Core Insights - Meta plans to launch a new model called "Avocado" in spring 2024, likely in a closed-source format, which has drawn inspiration from various open-source models, including Alibaba's Qwen [1][2] - The shift in Meta's strategy reflects a significant change in CEO Mark Zuckerberg's stance, moving from advocating for U.S. models to adopting insights from strong Chinese open-source models like Qwen [2][3] Group 1: Meta's New Model - Meta's new model "Avocado" is being developed by the newly formed TBD lab team and is expected to distill insights from multiple open-source models, including those from Google and OpenAI [1] - The launch of "Avocado" comes after Meta's Llama4 project faced setbacks, prompting a reevaluation of its approach to model development [2] Group 2: Alibaba's Qwen Model - Alibaba's Qwen model has seen significant success, with its derivative models expected to surpass Llama in number by August 2024 and in global download volume by October 2025 [2][3] - Qwen has been recognized for its performance, recently surpassing competitors like GPT-5 and Claude Opus 4, positioning itself among the top three models globally [3] Group 3: Industry Impact - Qwen's influence extends beyond Alibaba, with major companies like Amazon and Airbnb utilizing it for new business developments, and various institutions leveraging it for technological innovations [3][6] - The rapid growth of Qwen is evidenced by its monthly active users exceeding 30 million within just 23 days of its public testing, marking it as one of the fastest-growing AI applications globally [6]