{
  "slug": "muse-spark-open-816b-a42b",
  "order": 110,
  "name": "Muse Spark Open (816B-A42B)",
  "baseTitle": "Muse Spark Open 816B-A42B",
  "company": "Meta",
  "country": "United States",
  "flag": "🇺🇸",
  "date": "2026-10-15",
  "decoderType": "Sparse MoE",
  "attention": "Gated GQA with scaleless QK-Norm, local-only RoPE, global NoPE, and 3:1 sliding-window/global attention",
  "layerMix": "46 sliding-window + 16 global gated GQA",
  "scale": "816B total, 42B active (5.1% active)",
  "totalParams": 816000000000,
  "activeParams": 42000000000,
  "context": "1,048,576",
  "contextTokens": 1048576,
  "license": "Muse Model License",
  "kvCacheText": "248 KiB",
  "kvCacheKiB": 248,
  "kvCacheLevel": "High",
  "summary": "Meta's multimodal MoE model combines 62 latent-expert blocks with gated GQA, local-only RoPE, global NoPE, and learned residual gates; the figure focuses on the text decoder.",
  "keyDetail": "Selects 8 of 256 experts, compresses the expert path from 8,192 to 4,096 dimensions, normalizes each expert output before combining, and uses no shared expert.",
  "image": {
    "full": "https://sebastianraschka.com/llm-architecture-gallery/images/architectures/muse-spark-open.webp",
    "thumb": "https://sebastianraschka.com/llm-architecture-gallery/images/architectures/thumbnails/muse-spark-open.webp",
    "alt": "Figure 1: Muse Spark Open 816B-A42B",
    "width": 960,
    "height": 657
  },
  "articleUrl": null,
  "links": [
    {
      "label": "config.json",
      "href": "https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B/blob/main/config.json",
      "title": "meta-models/Muse-Spark-Open-816B-A42B"
    },
    {
      "label": "License",
      "href": "https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B/blob/main/LICENSE.md",
      "title": null
    }
  ],
  "concepts": [
    {
      "slug": "moe",
      "label": "MoE",
      "name": "Mixture of Experts",
      "url": "https://llm-atlas-7or.pages.dev/concepts/moe/"
    },
    {
      "slug": "latent-moe",
      "label": "LatentMoE",
      "name": "Latent MoE",
      "url": "https://llm-atlas-7or.pages.dev/concepts/latent-moe/"
    },
    {
      "slug": "gqa",
      "label": "GQA",
      "name": "Grouped-Query Attention",
      "url": "https://llm-atlas-7or.pages.dev/concepts/gqa/"
    },
    {
      "slug": "gated-attention",
      "label": "Gated Attention",
      "name": "Gated Attention",
      "url": "https://llm-atlas-7or.pages.dev/concepts/gated-attention/"
    },
    {
      "slug": "swa",
      "label": "SWA",
      "name": "Sliding Window Attention",
      "url": "https://llm-atlas-7or.pages.dev/concepts/swa/"
    },
    {
      "slug": "qk-norm",
      "label": "QK-Norm",
      "name": "QK-Norm",
      "url": "https://llm-atlas-7or.pages.dev/concepts/qk-norm/"
    },
    {
      "slug": "nope",
      "label": "NoPE",
      "name": "No Positional Embeddings",
      "url": "https://llm-atlas-7or.pages.dev/concepts/nope/"
    }
  ],
  "aai": {
    "v5": {
      "label": "AA Intelligence Index v5.0",
      "total": 24.7,
      "profile": {
        "Knowledge Work": 37,
        "Coding": 7.8,
        "Science": 12.1,
        "General": 41.7
      },
      "url": "https://artificialanalysis.ai/models/muse-spark-open"
    },
    "v4": {
      "label": "AA Intelligence Index v4.3.2",
      "total": 38.9,
      "profile": {
        "General": 45,
        "Scientific": 29.4,
        "Coding": 33.3,
        "Agents": 42.8
      },
      "url": "https://artificialanalysis.ai/models/muse-spark-open"
    }
  },
  "sourceUrl": "https://sebastianraschka.com/llm-architecture-gallery/#card-muse-spark-open-816b-a42b",
  "family": "moe",
  "url": "https://llm-atlas-7or.pages.dev/models/muse-spark-open-816b-a42b/",
  "markdownUrl": "https://llm-atlas-7or.pages.dev/models/muse-spark-open-816b-a42b.md"
}