Alibaba, ByteDance and Meta Ignite Personal Agent Race: No Table Without a Good Model

In late September, Alibaba announced at its Yunqi Conference that Qwen (Tongyi Qianwen) would accelerate building personal agents, and on October 13 it officially unveiled 'Qwen Go!', a personal assistant, starting limited invitation tests in the Qwen app. ByteDance is internally testing a personal agent project codenamed Spell alongside its Doubao products, and Tencent has TX Handy Bot and the Yuanbao assistant. The trend was triggered by Meta's Muse, an agent with its own cloud computer and browser that takes on daily tasks and long-term goals. OpenAI later launched dots. On the model side, on October 12 Meta released Muse Wonder and Muse Wonder Full, emphasizing Active Memory (retrieving useful information across conversations) and Active Focus (continuing toward goals after obstacles), alongside coding and professional capabilities. Separately, the Qwen team published QwenGyre, an online reinforcement learning framework for ultra-long-horizon agent tasks that run for hours and interact with environments hundreds of times. Submitted to arXiv on September 27, it aims to reduce idle GPU time and redundant computation. In Qwen's experiments, QwenGyre cut training time by about 44% versus an Async approach on the same GPU budget.