{
  "slug": "vibethinker-3b",
  "order": 82,
  "name": "VibeThinker-3B (3B)",
  "baseTitle": "VibeThinker-3B",
  "company": "Weibo",
  "country": "China",
  "flag": "🇨🇳",
  "date": "2026-06-15",
  "decoderType": "Dense",
  "attention": "8:1 GQA with RoPE",
  "layerMix": "36 GQA",
  "scale": "3B parameters",
  "totalParams": 3000000000,
  "activeParams": 3000000000,
  "context": "131,072",
  "contextTokens": 131072,
  "license": "MIT License",
  "kvCacheText": "36 KiB",
  "kvCacheKiB": 36,
  "kvCacheLevel": "Low",
  "summary": "Qwen2.5-Coder-based dense reasoner whose gains mainly come from verifier-driven post-training rather than a new block design.",
  "keyDetail": "Standard Qwen2.5-Coder-style dense stack strengthened by curriculum SFT, MGPO RL, Long2Short RL, offline self-distillation, and Instruct RL.",
  "image": {
    "full": "https://sebastianraschka.com/llm-architecture-gallery/images/architectures/vibethinker-3b.webp",
    "thumb": "https://sebastianraschka.com/llm-architecture-gallery/images/architectures/thumbnails/vibethinker-3b.webp",
    "alt": "Figure 1: VibeThinker-3B",
    "width": 921,
    "height": 960
  },
  "articleUrl": null,
  "links": [
    {
      "label": "config.json",
      "href": "https://huggingface.co/WeiboAI/VibeThinker-3B/blob/main/config.json",
      "title": "WeiboAI/VibeThinker-3B"
    },
    {
      "label": "License",
      "href": "https://huggingface.co/WeiboAI/VibeThinker-3B/blob/main/LICENSE",
      "title": null
    },
    {
      "label": "Tech report",
      "href": "https://arxiv.org/abs/2606.16140",
      "title": null
    }
  ],
  "concepts": [
    {
      "slug": "gqa",
      "label": "GQA",
      "name": "Grouped-Query Attention",
      "url": "https://llm-atlas-7or.pages.dev/concepts/gqa/"
    }
  ],
  "aai": null,
  "sourceUrl": "https://sebastianraschka.com/llm-architecture-gallery/#card-vibethinker-3b",
  "family": "dense",
  "url": "https://llm-atlas-7or.pages.dev/models/vibethinker-3b/",
  "markdownUrl": "https://llm-atlas-7or.pages.dev/models/vibethinker-3b.md"
}