Muse Wonder is a frontier reasoning model from Meta, designed for complex personal and professional agentic tasks. It accepts text, images, video, audio, and PDF documents and returns text, with a 1M-token context window.
The model handles long-horizon work: it plans, delegates to sub-agents, calls tools in parallel, and recovers from errors, so multi-step tasks run to completion without step-by-step direction. It can also operate browsers and desktop and mobile applications, and works across many files in existing repositories to make and verify changes. Muse Wonder carries important context across conversations and retrieves the details relevant to the task at hand. And it's trained for Active Focus, so it works toward your goal even when a task gets challenging.
In Meta's testing, it performs well on complex engineering tasks, extended agentic sessions, and multi-file refactors to tackle developers' most demanding work.
提供商
不同公司托管同一模型。OpenRouter 根据您选择的路由模式将请求路由到其中一家公司:Balanced(价格 + 速度)、Nitro(速度最快)或 Floor(价格最低)。
| US$4.00 | US$20.00 | US$0.20 | 1.88s | 103 tok/s | ||
| US$4.00 | US$20.00 | US$0.20 | 2.77s | 87 tok/s |
- 输入 US$4.00 / 输出 US$20.00103 tok/s99.96% 正常运行时间
- 输入 US$4.00 / 输出 US$20.0087 tok/s99.81% 正常运行时间
定价
客户为此模型实际支付的平均价格,与提供商公布的价格并列展示。由于缓存和折扣,实际支付的价格往往远低于标价。
加权平均值
Muse Wonder — 价格历史
| US$0.615 | US$20.00 | 89.1% | 87.5% | ||
| US$1.041 | US$20.00 | 77.9% | 12.5% |
- 有效输入价格 US$0.615 / 实际输出 US$20.00token 占比87.5%
- 有效输入价格 US$1.041 / 实际输出 US$20.00token 占比12.5%
性能
吞吐量是模型生成内容的速度(每秒 token 数,越高越好)。延迟是完整的往返时间(越低越好)。TTFT 是首个 token 的响应时间,即您看到内容出现前需要等待的时间(越低越好)。
P50,所有提供商中最佳
P50,最佳提供商
正常运行时间
正常运行时间是指过去 3 天中至少有一家提供商响应请求的时间百分比。可用性是指成功提供推理服务的时间百分比。OpenRouter 会持续监控,并在某家提供商返回错误时使用次优提供商。
过去 3 天的可用性
最近 72 小时过去 24 小时的可用性
当上游提供商发生错误时,如果您的请求筛选条件允许,我们可以通过路由到另一家健康的提供商来恢复服务。您可以通过以下接口以编程方式访问各提供商的正常运行时间数据: 端点 API。了解更多有关我们的 负载均衡和提供商路由选项。
基准测试
标准化评测得分,以及此模型针对共用基准测试提示词的实际输出。百分比越高越好,每项排名显示此模型在该来源所评测的模型中的位置。
| 来源 | 基准测试 | 得分 |
|---|---|---|
| Artificial Analysis | Muse Wonder (High) Intelligence Index | 40.5 |
| Artificial Analysis | Muse Wonder (High) Finance & Accounting Index | 50.1 |
| Artificial Analysis | Muse Wonder (High) Strategy & Ops Index | 54.1 |
| Artificial Analysis | Muse Wonder (High) Legal Index | 46.2 |
| Artificial Analysis | Muse Wonder (High) Engineering Index | 46.5 |
| Artificial Analysis | Muse Wonder (High) Economics Index | 47.9 |
| Artificial Analysis | Muse Wonder (High) AutomationBench-AA | 67.2% |
| Artificial Analysis | Muse Wonder (High) GDP.pdf | 29.2% |
| Artificial Analysis | Muse Wonder (High) HLE-Diamond | 34.5% |
| Artificial Analysis | Muse Wonder (High) Terminal-Bench 4.0 | 38.4% |
| Artificial Analysis | Muse Wonder (High) FrontierCode-AA 2.0 | 26.6% |
| Artificial Analysis | Muse Wonder (High) Terminal-Bench-Science | 21.9% |
| Artificial Analysis | Muse Wonder (High) CritPt v2 (Hard Subset) | 51.0% |
| Artificial Analysis | Muse Wonder (High) CritPt v2 | 70.9% |
| Artificial Analysis | Muse Wonder (High) AA-Omniscience Accuracy | 38.2% |
| Artificial Analysis | Muse Wonder (High) AA-Omniscience Non-Hallucination Rate | 77.8% |
| Artificial Analysis | Muse Wonder (High) MMMU-Pro | 83.8% |
| Artificial Analysis | Muse Wonder (Max) Intelligence Index | 43.6 |
| Artificial Analysis | Muse Wonder (Max) Finance & Accounting Index | 51.1 |
| Artificial Analysis | Muse Wonder (Max) Strategy & Ops Index | 54.5 |
| Artificial Analysis | Muse Wonder (Max) Legal Index | 46.9 |
| Artificial Analysis | Muse Wonder (Max) Healthcare & Medical Index | 48.7 |
| Artificial Analysis | Muse Wonder (Max) Engineering Index | 49.6 |
| Artificial Analysis | Muse Wonder (Max) Economics Index | 49.5 |
| Artificial Analysis | Muse Wonder (Max) AutomationBench-AA | 68.1% |
| Artificial Analysis | Muse Wonder (Max) GDP.pdf | 30.6% |
| Artificial Analysis | Muse Wonder (Max) Harvey LAB-AA | 10.0% |
| Artificial Analysis | Muse Wonder (Max) HLE | 50.1% |
| Artificial Analysis | Muse Wonder (Max) AA-LCR | 80.7% |
| Artificial Analysis | Muse Wonder (Max) GDPval-AA | 63.5% |
| Artificial Analysis | Muse Wonder (Max) CritPt | 28.3% |
| Artificial Analysis | Muse Wonder (Max) HLE-Diamond | 39.0% |
| Artificial Analysis | Muse Wonder (Max) SciCode | 60.3% |
| Artificial Analysis | Muse Wonder (Max) Terminal-Bench 4.0 | 52.0% |
| Artificial Analysis | Muse Wonder (Max) FrontierCode-AA 2.0 | 27.9% |
| Artificial Analysis | Muse Wonder (Max) FrontierCode 1.1 | 46.5% |
| Artificial Analysis | Muse Wonder (Max) Terminal-Bench-Science | 25.7% |
| Artificial Analysis | Muse Wonder (Max) CritPt v2 (Hard Subset) | 50.5% |
| Artificial Analysis | Muse Wonder (Max) CritPt v2 | 71.1% |
| Artificial Analysis | Muse Wonder (Max) AA-Omniscience Accuracy | 38.0% |
| Artificial Analysis | Muse Wonder (Max) AA-Omniscience Non-Hallucination Rate | 79.1% |
| Artificial Analysis | Muse Wonder (Max) MLCR-AA | 22.2% |
| Artificial Analysis | Muse Wonder (Max) VIALS-AA | 55.3% |
| Design Arena | Muse Wonder (max) Agents Arena Blender3d Elo | 1122 |
| Design Arena | Muse Wonder (max) Agents Arena Codetovideo Elo | 1216 |
| Design Arena | Muse Wonder (max) Agents Arena Htmlslides Elo | 1127 |
| Design Arena | Muse Wonder (max) Agents Arena Mobile Apps Elo | 1262 |
| Design Arena | Muse Wonder (max) Agents Arena Pdf Elo | 1226 |
| Design Arena | Muse Wonder (max) Agents Arena Python-Pptxslides Elo | 1033 |
| Design Arena | Muse Wonder (max) Agents Arena Spreadsheet Elo | 1232 |
| Design Arena | Muse Wonder (max) Models Arena 3D Elo | 1432 |
| Design Arena | Muse Wonder (max) Models Arena Code Categories Elo | 1391 |
| Design Arena | Muse Wonder (max) Models Arena Data Visualization Elo | 1282 |
| Design Arena | Muse Wonder (max) Models Arena Game Development Elo | 1412 |
| Design Arena | Muse Wonder (max) Models Arena UI Component Elo | 1360 |
| Design Arena | Muse Wonder (max) Models Arena Website Elo | 1377 |
| Vals AI | Wonder Access to Justice Benchmark | 73.0% |
| Vals AI | Wonder BioMysteryBench | 67.0% |
| Vals AI | Wonder Code Migration v1.1 | 72.4% |
| Vals AI | Wonder CyberBench v1.1 | 80.4% |
| Vals AI | Wonder EMB v1.1 | 72.5% |
| Vals AI | Wonder Finance Agent v2.1 | 65.5% |
| Vals AI | Wonder Vals Finance Index | 61.2% |
| Vals AI | Wonder Financial Coaching Bench | 73.8% |
| Vals AI | Wonder HLAB-Vals-1.1 | 23.8% |
| Vals AI | Wonder Immigration | 69.6% |
| Vals AI | Wonder IOI | 66.9% |
| Vals AI | Wonder Vals Legal Index | 75.3% |
| Vals AI | Wonder Legal Research Bench | 50.5% |
| Vals AI | Wonder MysteryMechanism | 27.5% |
| Vals AI | Wonder ProofBench v1.1 | 84.0% |
| Vals AI | Wonder Public Benefits Bench v1.1 | 70.2% |
| Vals AI | Wonder Vals Science Index | 39.1% |
| Vals AI | Wonder SWE-bench Multimodal | 45.6% |
| Vals AI | Wonder Tax Agent Bench | 36.3% |
| Vals AI | Wonder Terminal-Bench 4.0 | 49.5% |
| Vals AI | Wonder Terminal-Bench-Science | 22.9% |
| Vals AI | Wonder Vals Index | 64.9% |
| Vals AI | Wonder Vibe Code Bench 1-100 | 26.6% |
| Vals AI | Wonder Vibe Code Bench v1.1 | 83.1% |
应用
向此模型发送最多流量的公开应用。这能很好地反映实际生产工作负载的情况,也能提示此模型最适合哪些用例。
活动
此模型的 token 用量和请求流量随时间的变化。
API
可直接用于调用此模型的代码。OpenRouter 的 API 与 OpenAI 兼容——大多数 SDK 只需替换基础 URL 即可使用。切换模型时,唯一需要更改的是下方的模型标识符。
目前无法通过 EU 区域内路由使用。 区域内路由文档
常见问题
Muse Wonder is a frontier reasoning model from Meta, designed for complex personal and professional agentic tasks. It accepts text, images, video, audio, and PDF documents and returns text, with a 1M-token context window.
Muse Wonder 的费用为 US$4.00/百万输入 token 和 US$20.00/百万输出 token,Cache Read,价格为 US$0.20/M tokens 单独计费。
Muse Wonder 的上下文窗口为 1,048,576 token。
是的。Muse Wonder 接受 tools 和 tool_choice,以进行函数调用。 它还支持通过 response_format 中的 JSON schema 实现结构化输出。
Muse Wonder 接受 文本、图像、视频、PDF 等文件和音频 作为输入,并返回 文本。
在 OpenRouter 上,Muse Wonder 由 2 家提供商提供服务:Google Vertex和Meta。 请求会路由至最佳的可用提供商,并在故障时自动切换至其他提供商,您也可以通过提供商路由固定或排除提供商。
Muse Wonder 于 2026年10月12日 发布。
