Muse Wonder：面向个人与职业 AI 智能体的前沿模型
原题Muse Wonder: A Frontier Model For Personal & Professional AI Agents
推荐理由个人助手怎样记得住你的需求，又能坚持完成长任务？Meta 把主动记忆与持续专注放到同一模型框架里，适合产品团队拆分两类能力的评测方式。阅读时可区分官方已开放的模型、后续应用计划与厂商能力主张，设计贴近真实用户的试用任务。
读前速览
Meta 发布 Muse Wonder 与 Muse Wonder Full，称其在个人与专业智能体、记忆与编程基准上达到前沿水平，并以 Active Memory 与 Active Focus 为核心能力。
发布与可用性
Muse Wonder 通过 Meta Model API 与 Muse Code 提供，Muse Wonder Full 面向 Muse Code 订阅用户。
Active Memory 与 Active Focus
跨对话记住并主动调用用户上下文；持续推进困难目标并在关键操作前确认，用 FocusBench 与 SWE-Marathon 评估。
泛化能力
与 Muse Spark 1.3 对比，更强基座使后训练收益更好地迁移到分布外任务。
安全
称对 CBRN 与网络滥用的防护为迄今最强，完整安全报告稍后发布。
完整摘要
Meta 官方博客介绍 Muse Wonder 系列：Muse Wonder 今日起通过 Meta Model API 与 Muse Code 提供，Muse Wonder Full 向 Muse Code 订阅用户开放，后续将为 Muse 产品提供支持并优先向订阅用户推出。文章称，14 个月前 Meta 从第一性原理重建 AI 技术栈，新模型在吉瓦级 Prometheus 数据中心训练，在预训练与强化学习上延续 Muse Spark 的可预测扩展路线。重点能力包括：法律、金融、健康等领域专业知识（如 Harvey 的 Legal Agent Benchmark）；Active Memory，即跨对话记住与用户相关的重要上下文并在需要时主动调用；Active Focus，即持续推进困难目标、尝试替代方案、在关键操作前确认，用 FocusBench 与 SWE-Marathon 评估；以及前沿编程能力。文章还以 Muse Spark 1.3 对比说明，更强的基座让后训练收益更好地迁移到未针对训练的任务上；安全方面称其对 CBRN 与网络滥用的防护为迄今最强，完整安全报告将稍后发布。具体基准数字以图表呈现，正文未列出。
目录 · 9 节
在个人与专业智能体领域处于领先地位的前沿模型A frontier model that leads on personal and professional agents
今天，我们很高兴推出 Muse Wonder 系列，包括 Muse Wonder 和 Muse Wonder Full。这是我们能力最强的模型，在个人和专业智能体任务上均实现了最先进的性能。它们共同标志着我们沿扩展阶梯迈出的下一步，也是迈向个人超级智能的一大飞跃。Muse Wonder 将主动记忆与主动专注结合起来，持续关注对你重要的事，并勤勉地助你实现目标。Today, we’re excited to introduce the Muse Wonder series with Muse Wonder and Muse Wonder Full, our most capable models that deliver state-of-the-art performance on personal and professional agentic tasks. Together, they mark the next step on our scaling ladder and a major leap toward personal superintelligence. Muse Wonder brings together Active Memory and Active Focus to keep track of what matters to you and work diligently toward what you want to accomplish.
Muse Wonder is available today through the Meta Model API and Muse Code, and Muse Wonder Full is available today for Muse Code subscribers.
Muse Wonder 很快将为 Muse 中的体验提供支持，并率先向订阅用户推出。Muse Wonder 还将支持生成创作成果，以及需要更高智能的流程。未来，我们还将利用 Muse Wonder 开发专用模型，并将其提供给所有使用 Muse 的用户。Muse Wonder will soon power experiences in Muse, and will roll out first to subscribers. Muse Wonder will also power artifact generation and processes that require more intelligence. And in the future, we’ll be using Muse Wonder to develop specialized models that we can bring to everyone using Muse.
目录Contents
Scaling Toward Personal Superintelligence- 迈向个人超级智能的规模化发展
- 前沿能力
- 可泛化的能力
- 前沿领域的安全
迈向个人超级智能的规模化发展Scaling Toward Personal Superintelligence
十四个月前，我们开始从头彻底改革人工智能工作。在此期间，我们从基于基本原理重建技术栈，迈进到推动前沿发展。Fourteen months ago, we began a ground-up overhaul of our artificial intelligence efforts. In that time, we’ve gone from rebuilding our stack from first principles to pushing the frontier.
推出 Muse Spark 时，我们阐述了实现可预测、严谨规模化扩展的方法。今天发布的模型在预训练和强化学习方面延续了这项工作，打造出迄今能力最强的模型。Muse Wonder 和 Muse Wonder Full 在我们的吉瓦级 Prometheus 数据中心训练而成，汇聚了 Meta 各团队的成果，为工作和日常生活中的复杂任务带来前沿能力——这是我们朝着让数十亿产品用户拥有个人超级智能迈出的又一步。When we introduced Muse Spark, we outlined our approach to predictable, rigorous scaling. Today’s release builds on that work across pretraining and reinforcement learning, delivering our most capable models yet. Trained at our gigawatt-scale Prometheus data center, Muse Wonder and Muse Wonder Full bring together the work of teams across Meta to deliver frontier capabilities for complex tasks at work and in everyday life — another step toward bringing personal superintelligence to the billions of people who use our products.
前沿能力Frontier Capabilities
模型亮点Model Highlights
Muse Wonder 系列在个人和专业智能体、记忆及编程领域的关键基准测试中取得了最先进的成绩。The Muse Wonder series achieves state-of-the-art results across key benchmarks for personal and professional agents, memory, and coding.
以下完整结果将 Muse Wonder 和 Muse Wonder Full 与其他前沿模型在一套广泛的评测中进行了比较。The full results below compare Muse Wonder and Muse Wonder Full with other frontier models across a broad evaluation suite.
有关我们评估的更多详情，请参阅我们的报告。For more details about our evaluations, see our report.
以上结果全面展示了这些模型的性能。接下来，我们将重点深入考察其在真实任务中的四个核心领域：领域专业知识、主动记忆、主动专注和前沿编程。The results above provide a broad view of the models’ performance. We now take a closer look at four areas central to their work on real-world tasks: domain expertise, Active Memory, Active Focus, and frontier coding.
领域专业知识Domain Expertise
Muse Wonder 和 Muse Wonder Full 将专业知识应用于人们在工作和日常生活中遇到的问题，从理解租赁合同到读懂金融或健康信息。我们的评测涵盖法律、金融和健康领域的任务，用以检验这项能力。例如，Harvey 的法律智能体基准测试（LAB）要求模型处理源文件并遵循详细指示，生成法律工作成果。Muse Wonder and Muse Wonder Full apply specialized knowledge to problems people encounter at work and in daily life, from understanding a lease to making sense of financial or health information. Our evaluations include tasks in law, finance, and health that test this ability. Harvey’s Legal Agent Benchmark (LAB), for example, requires models to work with source documents and follow detailed instructions to produce legal work products.
活动内存Active Memory
迈向实用型个人和专业智能体，最关键的步骤之一是建立可靠的记忆。模型需要记住重要信息，并识别何时需要用到早先的信息。我们将其称为“主动记忆”：在不同对话中持续保留与你相关的重要背景信息，并主动检索与当前任务相关的细节。One of the most critical steps towards useful personal and professional agents is reliable memory. Models need to remember what matters and recognize when earlier information becomes relevant. We call this Active Memory: carrying important context about you across conversations and actively retrieving the details relevant to the task at hand.
当前焦点Active Focus
利用这些上下文完成任务，同样需要持续付出努力。「主动专注」描述了 Muse Wonder 如何持续朝着具有挑战性的目标前进：尝试不同的方法、克服阻碍，并根据新信息进行调整。在此过程中，它会在必要时请求澄清，并在采取重要行动前先征求确认。我们通过 FocusBench 评估专注力，通过 SWE-Marathon 评估持续执行能力；后者用于测试在长期软件项目中的工作表现。Using that context to complete a task also requires sustained effort. Active Focus describes how Muse Wonder continues working toward a challenging goal: trying alternative approaches, working through blockers, and adapting to new information. Along the way, it asks for clarification when needed and checks in before taking important actions. We evaluate focus through FocusBench and sustained execution through SWE-Marathon, which tests work on long-running software projects.
前沿编码Frontier Coding
编程将这些能力融为一体。开发软件需要跟踪需求、在整个代码库中进行修改，并反复测试和修订。Muse Wonder 和 Muse Wonder Full 带来前沿的编程性能，支持从小型应用到复杂软件的各类项目。我们使用多样化的长期任务训练这两个模型，帮助它们在整个过程中持续推进工作。Coding brings these capabilities together. Building software requires tracking requirements, making changes across a codebase, and working through repeated cycles of testing and revision. Muse Wonder and Muse Wonder Full deliver frontier coding performance, supporting projects from small applications to complex software. We trained both models on diverse, long-running tasks to help them sustain progress across this process.
Muse Wonder builds a playable 3D kitchen game from customizable image inputs.
可泛化的能力Capabilities That Generalize
两个模型的基准测试分数往往相近，但其中一个似乎更擅长处理种类广泛得多的陌生任务。这正是人们批评“刷榜”和称赞“强大模型气质”时所指的：能力覆盖面的广度，而少数领域里的热门基准测试无法充分体现这一点。Two models often have similar benchmark scores, yet one seems to handle a much wider range of unfamiliar tasks. That’s what complaints of “benchmaxxing” and praise for “big model smell” both get at: a breadth of capability that popular benchmarks in a few domains can’t fully capture.
为了量化这种不匹配，我们研究了随着测试时计算量增加，模型在后训练阶段重点针对的任务上的表现与陌生任务上的表现之间有何关系。下图比较了 Muse Spark 1.3 和 Muse Wonder Full，其中每个点表示一种不同推理力度下的表现。横轴显示模型在一个综合评分上的得分，该评分代表我们在后训练阶段重点关注的任务（“分布内”）。纵轴显示模型在一组后训练阶段未重点针对的任务上的平均得分（“分布外”）。To quantify this mismatch, we investigate how performance on tasks targeted during post-training relates to performance on unfamiliar tasks as we scale test-time compute. The plot below compares Muse Spark 1.3 and Muse Wonder Full, with each point denoting performance at a distinct reasoning effort. The horizontal axis shows the model’s score on a composite representative of the tasks we focused on during post-training (“in-distribution”). The vertical axis shows the average score on a set of tasks we did not target during post-training (“out-of-distribution”).
随着测试时计算量的增加，我们的两个模型在分布外任务上的提升幅度相近。但考虑到其分布内表现，Muse Wonder Full 在陌生任务上的表现甚至优于 Muse Spark 1.3 趋势所预测的水平。这一向上偏移源于 Muse Wonder 和 Muse Wonder Full 基于规模更大、能力更强的基础模型：更强的基础模型能带来更显著的后训练收益，并使这些收益更有效地迁移到陌生任务上。这种多面性对个人智能体尤为重要，因为它们需要应对目标、工具和约束条件的多种组合——而这些组合并非都能在训练期间预先设想到。Both of our models improve similarly on out-of-distribution tasks as we scale test-time compute. But given its in-distribution performance, Muse Wonder Full performs even better on unfamiliar tasks than the Muse Spark 1.3 trend would predict. This upward shift is because Muse Wonder and Muse Wonder Full build on larger, stronger base models: a stronger base yields post-training gains that carry over more effectively to unfamiliar tasks. That versatility especially matters for personal agents, which need to handle many combinations of goals, tools, and constraints — not all of which we can anticipate during training.
Working directly in a browser and across apps, Muse Wonder builds a running plan (Map data © OpenStreetMap contributors) based on personal preferences and circumstances.
前沿领域的安全保障Safety at the Frontier
要让每个人都能拥有个人超级智能，需要在对齐和安全方面进行大量投入。个人智能体会代表用户采取行动，也可能处理消息、日历和私人文档等上下文信息。这种访问权限要求模型能够谨慎处理私人信息，并抵御滥用。我们训练了 Muse Wonder 系列模型，使其在这两个方面都处于领先地位，既能抵御对抗性指令，也能谨慎处理私人信息。Delivering personal superintelligence to everyone requires a significant investment in alignment and safety. A personal agent takes actions on a user’s behalf and might handle context such as messages, calendars, and private documents. That level of access demands a model that exercises discretion with private information and resists misuse. We trained the Muse Wonder family to lead on both dimensions, with robustness to adversarial instructions and careful handling of private information.
随着模型能力不断增强，我们的目标是帮助人们安全、负责任地运用这些能力。根据我们的《Meta 超级智能扩展框架》，Muse Wonder 迄今提供了最强有力的防护措施，以防范 CBRN（化学、生物、放射性和核）及网络技术滥用，同时保留防御性用途所需的前沿网络能力。As models become more capable, our goal is to help people harness those capabilities safely and responsibly. Muse Wonder offers our strongest safeguards to date against CBRN (chemical, biological, radiological, and nuclear) and cyber misuse, while preserving access to frontier cyber capabilities for defensive use, per our Meta Superintelligence Scaling Framework.
Muse Wonder 是迄今为止与我们的理念最契合的模型，而对齐是我们构建个人超级智能的核心。包含更多安全评估的完整报告即将发布。Muse Wonder is our most aligned model to date, and alignment is central to how we build personal superintelligence. A full report with additional safety evaluations will be available soon.
我们希望 Muse Wonder 和 Muse Wonder Full 能赋予更多人追逐梦想、构建理想世界的能力。这是迈向人人都能拥有个人超级智能的重要技术一步，未来还有更多精彩。We hope that Muse Wonder and Muse Wonder Full give more people the power to pursue their dreams and build the world they want. This is a major technical step toward personal superintelligence for everyone, with much more to come.