Meta launches Muse Wonder and Wonder Full, scoring above Gemini 4 Argon
Meta Superintelligence Lab released Muse Wonder and Wonder Full, which score 52/56 on the Artificial Analysis index, ahead of GPT-6 Astra and Gemini 4 Argon, and feature Active Memory and Active Focus for personal and professional agents.

Meta Superintelligence Lab 发布 Muse Wonder 与 Wonder Full,Artificial Analysis 指数 52/56, 超 GPT-6 Astra 与 Gemini 4 Argon, 仅落后 Claude Opus 5.5 两分 这是 Meta 迄今为止规模最大的前沿模型,专为个人及专业 AI 智能体(Agent)优化,在法律、金融、医疗和复杂编程领域表现卓越。 该模型在 Meta 千兆瓦级(gigawatt-scale)Prometheus 数据中心进行训练,具备“主动记忆”(Active Memory)功能以跨对话保持长期上下文,以及“主动聚焦”(Active Focus)功能,能够通过自适应规划持续推进目标实现。 Muse Wonder 在 Harvey Legal Agent Benchmark 和 SWE-Marathon 等智能体基准测试中取得了业界领先的成绩; 得益于扩展的推理时计算(test-time compute)能力以及对新颖任务的泛化能力,其表现优于以往模型。
This post quotes another post: “Meta releases Muse Wonder frontier model”