{
  "slug": "ouro-thinking-2-6b",
  "order": 52,
  "name": "Ouro-Thinking (2.6B)",
  "baseTitle": "Ouro-Thinking 2.6B",
  "company": "ByteDance",
  "country": "China",
  "flag": "🇨🇳",
  "date": "2025-10-29",
  "decoderType": "Dense",
  "attention": "MHA with RoPE",
  "layerMix": "48 shared MHA layers × 4 passes",
  "scale": "2.6B parameters",
  "totalParams": 2600000000,
  "activeParams": 2600000000,
  "context": "65,536",
  "contextTokens": 65536,
  "license": "Apache License 2.0",
  "kvCacheText": "1.5 MiB",
  "kvCacheKiB": 1536,
  "kvCacheLevel": "Very high",
  "summary": "ByteDance's dense LoopLM reuses one 48-layer MHA stack across four recurrent passes and trains an exit gate over the intermediate representations.",
  "keyDetail": "Reuses the same 48-layer MHA stack across four passes, yielding 192 block applications, with sandwich RMSNorm and a learned exit gate.",
  "image": {
    "full": "https://sebastianraschka.com/llm-architecture-gallery/images/architectures/ouro-2-6b.webp",
    "thumb": "https://sebastianraschka.com/llm-architecture-gallery/images/architectures/thumbnails/ouro-2-6b.webp",
    "alt": "Figure 1: Ouro-Thinking 2.6B",
    "width": 960,
    "height": 878
  },
  "articleUrl": null,
  "links": [
    {
      "label": "config.json",
      "href": "https://huggingface.co/ByteDance/Ouro-2.6B-Thinking/blob/main/config.json",
      "title": "ByteDance/Ouro-2.6B-Thinking"
    },
    {
      "label": "License",
      "href": "https://huggingface.co/ByteDance/Ouro-2.6B-Thinking/blob/main/README.md#license",
      "title": null
    },
    {
      "label": "Tech report",
      "href": "https://arxiv.org/abs/2510.25741",
      "title": null
    }
  ],
  "concepts": [
    {
      "slug": "looped-depth-sharing",
      "label": "Looped Transformer",
      "name": "Looped Transformer",
      "url": "https://llm-atlas-7or.pages.dev/concepts/looped-depth-sharing/"
    },
    {
      "slug": "mha",
      "label": "MHA",
      "name": "Multi-Head Attention",
      "url": "https://llm-atlas-7or.pages.dev/concepts/mha/"
    }
  ],
  "aai": null,
  "sourceUrl": "https://sebastianraschka.com/llm-architecture-gallery/#card-ouro-thinking-2-6b",
  "family": "dense",
  "url": "https://llm-atlas-7or.pages.dev/models/ouro-thinking-2-6b/",
  "markdownUrl": "https://llm-atlas-7or.pages.dev/models/ouro-thinking-2-6b.md"
}