{
  "id": "CVE-2026-116816",
  "url": "https://poc.griffinguard.io/cve/CVE-2026-116816",
  "published_at": "2026-10-27T00:33:58.581Z",
  "title": "llama.cpp b5899 before b10261 Heap Out-of-Bounds Write via plamo2 Tokenizer",
  "vendor": "ggml-org",
  "product": "llama.cpp",
  "cvss": {
    "score": 8.5,
    "source": "cna",
    "vector": "CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N"
  },
  "kev": {
    "in_kev": false,
    "added_at": null,
    "days_after_publication": null
  },
  "first_poc": null,
  "poc_repo_count": 0,
  "exploit_count": 0,
  "suspicious_repo_count": 0,
  "edge_device": null,
  "description": "llama.cpp b5899 before b10261 contains a heap out-of-bounds write vulnerability in llm_tokenizer_plamo2::build that allows attackers to corrupt heap memory via crafted GGUF models. Attackers can supply BYTE-typed token text like \u003c0x-1\u003e whose negative parsed value writes token ids before the bytes_ buffer during model loading, causing crashes or potential code execution.",
  "cwe": "CWE-129",
  "cvss_by_source": {
    "cna": 8.5
  },
  "repos": [],
  "remediation": {
    "affected": [
      {
        "vendor": "ggml-org",
        "product": "llama.cpp",
        "ranges": [
          "\u003e= b5899, \u003c b10261"
        ],
        "fixed_in": [
          "b10261"
        ]
      }
    ],
    "upgrade": [
      "Upgrade llama.cpp to b10261 or later."
    ],
    "references": [
      {
        "kind": "Patch",
        "url": "https://github.com/ggml-org/llama.cpp/pull/25969"
      },
      {
        "kind": "Patch",
        "url": "https://github.com/ggml-org/llama.cpp/pull/26511"
      },
      {
        "kind": "Patch",
        "url": "https://github.com/ggml-org/llama.cpp/commit/cf06ad7dfe79e7bf2934150d4fefe1ce8db05490"
      },
      {
        "kind": "Release notes",
        "url": "https://github.com/ggml-org/llama.cpp/releases/tag/b10261"
      }
    ]
  },
  "exploit_refs": [],
  "hunt_profile": [],
  "rules": []
}
