LecturaInvestigaciónRadarMarco de inversión
Iniciar sesión / Registrarse
← Volver a la lectura
TemaEditorial de DisconfirmAIActualizado 2026-11-06 23:14

Meta Muse模型

3 pasos2 piezas4 Fuentes

Meta超级智能实验室(MSL)发布Muse Wonder系列模型,标志着其正式进入前沿AI竞争领域。该系列包含面向API用户的Muse Wonder和仅限Muse Code订阅者的Muse Wonder Full。第三方评测显示,Wonder Full在智能指数上得分56,接近Claude Opus 5.5的58分,并在长程编码和法律代理任务中表现强劲,但在通用基准上仍略逊一筹。Meta强调其“主动记忆”和“主动聚焦”能力,旨在支持个人和专业代理工作。此外,Meta宣布向加州理工等五所美国大学提供模型与资金,以加速医学与科学突破,展示了其在科研领域的落地应用。

Cómo llegó hasta aquí

  1. 2026-10-12

    Meta Superintelligence Labs发布Muse Wonder和Muse Wonder Full模型。Muse Wonder通过Meta Model API和Muse Code提供,Muse Wonder Full仅对Muse Code订阅者开放。定价为每百万输入token $4,输出token $20。

    latent.space ↗
  2. 2026-10-12

    Artificial Analysis发布独立评测报告,指出Muse Wonder Full在Intelligence Index v4.3中得分为56,Muse Wonder为52,均显著高于前代Muse Spark 1.3(48),但略低于Claude Opus 5.5(58)。

    artificialanalysis.ai ↗
  3. 2026-10-28

    Meta宣布支持包括加州理工学院、俄亥俄州立大学医学院在内的五所美国大学,利用Muse模型及资金加速医学与科学研究,重点解决心脏衰竭预测、数学猜想证明等难题。

    research.meta.ai ↗

Lo que aún está pendiente

  • Muse Wonder Full在通用基准上与Claude Opus 5.5的差距是否会在后续版本中缩小?
  • Meta与五所大学的合作具体产出了哪些可验证的科学或医学突破?
  • 关于“基准测试作弊”的行业争议是否会影响Muse Wonder的市场信任度?

Fuentes

Iniciar sesión / Registrarse
LecturaInvestigaciónRadarMarco de inversión
latent.space
Latent Space AINews
↗
  • artificialanalysis.aiArtificial Analysis↗
  • research.meta.aiMeta AI Research Blog↗
  • techcrunch.comTechCrunch↗
  • Cobertura en este tema

    • Meta Announces Support for Five US Universities to Accelerate Medical and Scientific Breakthroughs with Muse Models2026-10-29Noticias

      Meta provides AI models, funding, and collaboration to Caltech, OSU, LSU, IU, and UTHealth to tackle specific challenges like heart failure prediction and mathematical conjectures.

    • Latent Space AINews: Meta Launches Muse Wonder Series, Third-Party Evals Show Near-Claude Opus 5.5 Performance2026-10-23Noticias

      This AINews episode focuses on the release of Muse Wonder and Wonder Full by Meta Superintelligence Labs (MSL). The program aggregates independent scores from Artificial Analysis, developer feedback, and pricing strategies, noting that while Wonder Full performs strongly in long-horizon coding and legal agent tasks, it still trails Claude Opus 5.5 on general benchmarks, alongside discussions on 'benchmark gaming' controversies.