{
  "slug": "ornith-1-5-35b-a3b",
  "order": 98,
  "name": "Ornith 1.5 (35B-A3B)",
  "baseTitle": "Ornith 1.5 35B-A3B",
  "company": "Ornith",
  "country": "United States",
  "flag": "🇺🇸",
  "date": "2026-08-23",
  "decoderType": "Sparse hybrid",
  "attention": "3:1 Gated DeltaNet and Gated Attention",
  "layerMix": "10 gated attention + 30 DeltaNet",
  "scale": "35B total, 3B active (8.6% active)",
  "totalParams": 35000000000,
  "activeParams": 3000000000,
  "context": "262,144",
  "contextTokens": 262144,
  "license": "MIT License",
  "kvCacheText": "20 KiB",
  "kvCacheKiB": 20,
  "kvCacheLevel": "Very low",
  "summary": "Uses the same sparse-hybrid decoder geometry as Qwen3.6 35B-A3B, paired with Ornith's self-improvement training recipe.",
  "keyDetail": "Matches Qwen3.6 35B-A3B's 40-layer decoder geometry, with 256 routed experts, top-8 routing, and one shared expert.",
  "image": {
    "full": "https://sebastianraschka.com/llm-architecture-gallery/images/architectures/ornith-1-5-35b-a3b.webp",
    "thumb": "https://sebastianraschka.com/llm-architecture-gallery/images/architectures/thumbnails/ornith-1-5-35b-a3b.webp",
    "alt": "Figure 1: Ornith 1.5 35B-A3B",
    "width": 960,
    "height": 904
  },
  "articleUrl": null,
  "links": [
    {
      "label": "config.json",
      "href": "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B/blob/main/config.json",
      "title": "ornith-ai/Ornith-1.5-35B-A3B"
    },
    {
      "label": "Tech report",
      "href": "https://ornith.ai/ornith_1_5.html",
      "title": null
    }
  ],
  "concepts": [
    {
      "slug": "hybrid-attention",
      "label": "Hybrid Attention",
      "name": "Hybrid Attention",
      "url": "https://llm-atlas-7or.pages.dev/concepts/hybrid-attention/"
    },
    {
      "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": "qk-norm",
      "label": "QK-Norm",
      "name": "QK-Norm",
      "url": "https://llm-atlas-7or.pages.dev/concepts/qk-norm/"
    },
    {
      "slug": "shortconv",
      "label": "ShortConv",
      "name": "Short Convolution",
      "url": "https://llm-atlas-7or.pages.dev/concepts/shortconv/"
    },
    {
      "slug": "moe",
      "label": "MoE",
      "name": "Mixture of Experts",
      "url": "https://llm-atlas-7or.pages.dev/concepts/moe/"
    },
    {
      "slug": "mtp",
      "label": "MTP",
      "name": "Multi-Token Prediction",
      "url": "https://llm-atlas-7or.pages.dev/concepts/mtp/"
    }
  ],
  "aai": null,
  "sourceUrl": "https://sebastianraschka.com/llm-architecture-gallery/#card-ornith-1-5-35b-a3b",
  "family": "hybrid-moe",
  "url": "https://llm-atlas-7or.pages.dev/models/ornith-1-5-35b-a3b/",
  "markdownUrl": "https://llm-atlas-7or.pages.dev/models/ornith-1-5-35b-a3b.md"
}