{
  "story_id": "0602bc0e4fe12ddd7b80d8f9980f166c",
  "desk": "drm3",
  "revision": 1,
  "published_at": "2026-09-05T09:20:36.000Z",
  "content_hash": "176a01bed8636393687eac31837a30f7c46469a657db238ec1b0d4037ccfce68",
  "hash_basis": "sha256 over `headline\\ndek\\nprose`, plus `\\n` + the canonical citations JSON when any source is placed, plus `\\n#blog` for blogs",
  "basis": {
    "headline": "IFM's K2 Horizon AI Model Launches with Reward-Hacking Audit",
    "dek": "IFM launches K2 Horizon, the largest open-source AI drop, featuring a 7B model and a reward-hacking audit.",
    "prose": "IFM launched K2 Horizon, which is described as the largest open-source drop in AI history. [^1]\n\nIFM announced the release of K2 Horizon, a fleet of six open-source AI models (375B-A23B, 36B-A4B, 32B, 7B, 3.7B, and 0.9B) designed to span deployment from edge devices to enterprise. [^2]\n\nA reward-hacking audit was conducted on the models as part of the launch process. [^3]\n\nThe K2 Horizon release features a 7B parameter model that is highlighted as standing out among the offerings. [^4]\n\nIFM claims the 0.9B, 3.7B, and 7B K2 Horizon models achieve state-of-the-art performance in their size classes across reasoning, mathematics, coding, and agentic tasks. [^5]\n\nIFM states that K2 Horizon is its most comprehensive open release to date, opening the full training lifecycle from pretraining through agentic post-training, including intermediate checkpoints, training data or recipes, code, configurations, and logs. [^6]\n\nThe article notes a specific discrepancy of 3.37 points related to the model's performance or metrics. [^7]\n\nIFM claims the K2 Horizon 0.9B model achieves an AIME 2026 score above 48, along with strong reasoning, tool-use, and agentic capabilities. [^8]",
    "cited": "[{\"statement\":\"IFM launched K2 Horizon, which is described as the largest open-source drop in AI history.\",\"source\":\"medium.com\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-05T09:20:36.000Z\",\"publisher_count\":1,\"sources\":[\"medium.com\"]},{\"statement\":\"IFM announced the release of K2 Horizon, a fleet of six open-source AI models (375B-A23B, 36B-A4B, 32B, 7B, 3.7B, and 0.9B) designed to span deployment from edge devices to enterprise.\",\"source\":\"Institute of Foundation Models\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T15:36:43.000Z\",\"publisher_count\":1,\"sources\":[\"Institute of Foundation Models\"]},{\"statement\":\"A reward-hacking audit was conducted on the models as part of the launch process.\",\"source\":\"medium.com\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-05T09:20:36.000Z\",\"publisher_count\":1,\"sources\":[\"medium.com\"]},{\"statement\":\"The K2 Horizon release features a 7B parameter model that is highlighted as standing out among the offerings.\",\"source\":\"medium.com\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-05T09:20:36.000Z\",\"publisher_count\":1,\"sources\":[\"medium.com\"]},{\"statement\":\"IFM claims the 0.9B, 3.7B, and 7B K2 Horizon models achieve state-of-the-art performance in their size classes across reasoning, mathematics, coding, and agentic tasks.\",\"source\":\"Institute of Foundation Models\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T15:36:43.000Z\",\"publisher_count\":1,\"sources\":[\"Institute of Foundation Models\"]},{\"statement\":\"IFM states that K2 Horizon is its most comprehensive open release to date, opening the full training lifecycle from pretraining through agentic post-training, including intermediate checkpoints, training data or recipes, code, configurations, and logs.\",\"source\":\"Institute of Foundation Models\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T15:36:43.000Z\",\"publisher_count\":1,\"sources\":[\"Institute of Foundation Models\"]},{\"statement\":\"The article notes a specific discrepancy of 3.37 points related to the model's performance or metrics.\",\"source\":\"medium.com\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-05T09:20:36.000Z\",\"publisher_count\":1,\"sources\":[\"medium.com\"]},{\"statement\":\"IFM claims the K2 Horizon 0.9B model achieves an AIME 2026 score above 48, along with strong reasoning, tool-use, and agentic capabilities.\",\"source\":\"Institute of Foundation Models\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T15:36:43.000Z\",\"publisher_count\":1,\"sources\":[\"Institute of Foundation Models\"]}]",
    "kind": "news"
  },
  "receipt_verify": "Ed25519 over the dot-joined string `slice_hash.cursor_from.cursor_to.view.view_version.row_count`; public_key and sig are base64url of the raw 32-byte key / 64-byte signature",
  "receipt": null,
  "receipt_note": "this revision predates receipt-keeping (before v0.37.0); the filed row lives in the record",
  "generation_chain": {
    "wire": {
      "stream": "fountain_news",
      "story_id": "eaf38bc396bc3a69f1e01def5991268b",
      "thread_id": "4f1e56f4001e9c3b1a9d28f1bed96b52",
      "thread_label": "K2 Horizon",
      "novelty": "UPDATE",
      "content_hash": "9a710b14dea6fdffefcbe87fd37c45ac220d029a067ca2e58853fa2a6c49a338",
      "last_published_at": "2026-09-05T09:20:36.000Z",
      "read_receipt": {
        "slice_hash": "fbfb6f8cde02f621256660fe75c7a845a4a320cd83f26c76238a8b684f3dc6e8",
        "cursor_from": "eyJ0cyI6IjIwMjYtMDktMDVUMDg6MjM6MDIuMDAwMDAwWiIsImlkIjoiMjdkYjdmYzZiMWNhY2UzOWZlOWU2ZGE5YzhiOTJiMTgiLCJ2IjoiMSJ9",
        "cursor_to": "eyJ0cyI6IjIwMjYtMDktMDVUMDk6MjU6MzguMDAwMDAwWiIsImlkIjoiM2UzODQ5MGM1YWVhNDA1ZThmMzNmMjY3ODNjMWQzZmYiLCJ2IjoiMSJ9",
        "view": "v_fountain_news",
        "view_version": "1",
        "row_count": 100,
        "window_days": 3,
        "bytes_scanned": 10804931,
        "credits": 8,
        "price_per_100_rows": 8,
        "sig": "-CR8xsszMJbIK3t22jNXY2QYxGXJX1GPmy4KhIV8zU-SSJoe4y8qHjqOw8b88fakyJgBZZvpLBlqqtFpoHWVCQ",
        "public_key": "bMUigy8O0jOnBxQ4Sc-5lwhIZ8LQVAhxMbR7qESVuUE",
        "signer_path": "lakehouse/data-extract/v1",
        "alg": "Ed25519",
        "signed": true
      }
    },
    "written_at": "2026-09-05T12:29:30.570Z"
  },
  "cited_facts": [
    {
      "statement": "IFM launched K2 Horizon, which is described as the largest open-source drop in AI history.",
      "source": "medium.com",
      "instrument": "News",
      "claim_key": null,
      "published_at": "2026-09-05T09:20:36.000Z",
      "publisher_count": 1,
      "sources": [
        "medium.com"
      ]
    },
    {
      "statement": "IFM announced the release of K2 Horizon, a fleet of six open-source AI models (375B-A23B, 36B-A4B, 32B, 7B, 3.7B, and 0.9B) designed to span deployment from edge devices to enterprise.",
      "source": "Institute of Foundation Models",
      "instrument": "News",
      "claim_key": null,
      "published_at": "2026-09-03T15:36:43.000Z",
      "publisher_count": 1,
      "sources": [
        "Institute of Foundation Models"
      ]
    },
    {
      "statement": "A reward-hacking audit was conducted on the models as part of the launch process.",
      "source": "medium.com",
      "instrument": "News",
      "claim_key": null,
      "published_at": "2026-09-05T09:20:36.000Z",
      "publisher_count": 1,
      "sources": [
        "medium.com"
      ]
    },
    {
      "statement": "The K2 Horizon release features a 7B parameter model that is highlighted as standing out among the offerings.",
      "source": "medium.com",
      "instrument": "News",
      "claim_key": null,
      "published_at": "2026-09-05T09:20:36.000Z",
      "publisher_count": 1,
      "sources": [
        "medium.com"
      ]
    },
    {
      "statement": "IFM claims the 0.9B, 3.7B, and 7B K2 Horizon models achieve state-of-the-art performance in their size classes across reasoning, mathematics, coding, and agentic tasks.",
      "source": "Institute of Foundation Models",
      "instrument": "News",
      "claim_key": null,
      "published_at": "2026-09-03T15:36:43.000Z",
      "publisher_count": 1,
      "sources": [
        "Institute of Foundation Models"
      ]
    },
    {
      "statement": "IFM states that K2 Horizon is its most comprehensive open release to date, opening the full training lifecycle from pretraining through agentic post-training, including intermediate checkpoints, training data or recipes, code, configurations, and logs.",
      "source": "Institute of Foundation Models",
      "instrument": "News",
      "claim_key": null,
      "published_at": "2026-09-03T15:36:43.000Z",
      "publisher_count": 1,
      "sources": [
        "Institute of Foundation Models"
      ]
    },
    {
      "statement": "The article notes a specific discrepancy of 3.37 points related to the model's performance or metrics.",
      "source": "medium.com",
      "instrument": "News",
      "claim_key": null,
      "published_at": "2026-09-05T09:20:36.000Z",
      "publisher_count": 1,
      "sources": [
        "medium.com"
      ]
    },
    {
      "statement": "IFM claims the K2 Horizon 0.9B model achieves an AIME 2026 score above 48, along with strong reasoning, tool-use, and agentic capabilities.",
      "source": "Institute of Foundation Models",
      "instrument": "News",
      "claim_key": null,
      "published_at": "2026-09-03T15:36:43.000Z",
      "publisher_count": 1,
      "sources": [
        "Institute of Foundation Models"
      ]
    }
  ],
  "note": "A signature proves who filed this and that it has not changed since. It never makes a claim true."
}