{
  "slug": "llama-4-maverick",
  "order": 8,
  "name": "Llama 4 Maverick (400B)",
  "baseTitle": "Llama 4 Maverick",
  "company": "Meta",
  "country": "United States",
  "flag": "🇺🇸",
  "date": "2025-04-05",
  "decoderType": "Sparse MoE",
  "attention": "GQA",
  "layerMix": "36 chunked + 12 full GQA",
  "scale": "400B total, 17B active (4.3% active)",
  "totalParams": 400000000000,
  "activeParams": 17000000000,
  "context": "1,000,000",
  "contextTokens": 1000000,
  "license": "Llama 4 Community License Agreement",
  "kvCacheText": "192 KiB",
  "kvCacheKiB": 192,
  "kvCacheLevel": "High",
  "summary": "Meta's large MoE follows the DeepSeek V3 playbook but with a more conventional attention stack.",
  "keyDetail": "Alternates dense and MoE blocks and uses fewer, larger experts than DeepSeek V3.",
  "image": {
    "full": "https://sebastianraschka.com/llm-architecture-gallery/images/architectures/llama-4-maverick-400b.webp",
    "thumb": "https://sebastianraschka.com/llm-architecture-gallery/images/architectures/thumbnails/llama-4-maverick-400b.webp",
    "alt": "Figure 17: Llama 4 Maverick",
    "width": 960,
    "height": 932
  },
  "articleUrl": "https://magazine.sebastianraschka.com/p/the-big-llm-architecture-comparison#%C2%A75-llama-4",
  "links": [
    {
      "label": "View in article",
      "href": "https://magazine.sebastianraschka.com/p/the-big-llm-architecture-comparison#%C2%A75-llama-4",
      "title": null
    },
    {
      "label": "config.json",
      "href": "https://huggingface.co/meta-llama/Llama-4-Maverick-17B-128E-Instruct/blob/main/config.json",
      "title": "meta-llama/Llama-4-Maverick-17B-128E-Instruct"
    },
    {
      "label": "Tech report",
      "href": "https://ai.meta.com/blog/llama-4-multimodal-intelligence/",
      "title": null
    }
  ],
  "concepts": [
    {
      "slug": "gqa",
      "label": "GQA",
      "name": "Grouped-Query Attention",
      "url": "https://llm-atlas-7or.pages.dev/concepts/gqa/"
    },
    {
      "slug": "moe",
      "label": "MoE",
      "name": "Mixture of Experts",
      "url": "https://llm-atlas-7or.pages.dev/concepts/moe/"
    }
  ],
  "aai": {
    "v4": {
      "label": "AA Intelligence Index v4.3.2",
      "total": 9.3,
      "profile": {
        "General": 18.8,
        "Scientific": 2.5,
        "Coding": 15.9,
        "Agents": 0
      },
      "url": "https://artificialanalysis.ai/models/llama-4-maverick"
    }
  },
  "sourceUrl": "https://sebastianraschka.com/llm-architecture-gallery/#card-llama-4-maverick",
  "family": "moe",
  "url": "https://llm-atlas-7or.pages.dev/models/llama-4-maverick/",
  "markdownUrl": "https://llm-atlas-7or.pages.dev/models/llama-4-maverick.md"
}