{
  "slug": "step-5-preview-600b",
  "order": 103,
  "name": "Step-5 Preview (600B)",
  "baseTitle": "Step-5 Preview 600B",
  "company": "StepFun",
  "country": "China",
  "flag": "🇨🇳",
  "date": "2026-09-20",
  "decoderType": "Sparse MoE",
  "attention": "Gated GQA with QK-Norm and 3:1 sliding-window/sparse attention",
  "layerMix": "69 sliding-window + 23 sparse GQA + 3 MTP layers",
  "scale": "600B total, 27B active (4.5% active)",
  "totalParams": 600000000000,
  "activeParams": 27000000000,
  "context": "1,048,576",
  "contextTokens": 1048576,
  "license": "N/A",
  "kvCacheText": null,
  "kvCacheKiB": null,
  "kvCacheLevel": null,
  "summary": "StepFun's 92-layer MoE pairs 512-token sliding-window attention with sparse GQA for a 1M-token context window; the figure shows the text decoder.",
  "keyDetail": "Uses top-512 sparse indexing with block size 8, 352 routed experts with 8 active plus one shared expert, and MTP-3. The first three and final backbone layers use dense FFNs with intermediate size 13,824.",
  "image": {
    "full": "https://sebastianraschka.com/llm-architecture-gallery/images/architectures/step-5.webp",
    "thumb": "https://sebastianraschka.com/llm-architecture-gallery/images/architectures/thumbnails/step-5.webp",
    "alt": "Figure 1: Step-5 Preview 600B",
    "width": 960,
    "height": 940
  },
  "articleUrl": null,
  "links": [
    {
      "label": "config.json (mirror)",
      "href": "https://huggingface.co/SHSLab/Step-5-Preview-BF16/blob/main/config.json",
      "title": "SHSLab/Step-5-Preview-BF16"
    },
    {
      "label": "Tech report",
      "href": "https://www.stepfun.com/step-5-preview",
      "title": null
    }
  ],
  "concepts": [
    {
      "slug": "gqa",
      "label": "GQA",
      "name": "Grouped-Query Attention",
      "url": "https://llm-atlas-7or.pages.dev/concepts/gqa/"
    },
    {
      "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": "gated-attention",
      "label": "Gated Attention",
      "name": "Gated Attention",
      "url": "https://llm-atlas-7or.pages.dev/concepts/gated-attention/"
    },
    {
      "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": {
    "v5": {
      "label": "AA Intelligence Index v5.0",
      "total": 28.5,
      "profile": {
        "Knowledge Work": 40.9,
        "Coding": 20.5,
        "Science": 14.4,
        "General": 38
      },
      "url": "https://artificialanalysis.ai/models/step-5"
    },
    "v4": {
      "label": "AA Intelligence Index v4.3.2",
      "total": 43.7,
      "profile": {
        "General": 43,
        "Scientific": 33.7,
        "Coding": 46.1,
        "Agents": 49.7
      },
      "url": "https://artificialanalysis.ai/models/step-5"
    }
  },
  "sourceUrl": "https://sebastianraschka.com/llm-architecture-gallery/#card-step-5-preview-600b",
  "family": "moe",
  "url": "https://llm-atlas-7or.pages.dev/models/step-5-preview-600b/",
  "markdownUrl": "https://llm-atlas-7or.pages.dev/models/step-5-preview-600b.md"
}