Skip to content
Heist/ feed
Back to feed
MediumFix available

vLLM: Qwen2-VL / Qwen3-VL video samplers bound on request-controlled max_frames, which the num_frames ceiling does not reach

Type: advisorySources: OSV2 source records

### Summary An unauthenticated remote attacker can exhaust the memory of the vLLM API-server process by raising the request-level `media_io_kwargs.video.max_frames` and `fps` knobs on any deployment serving a Qwen2-VL or Qwen3-VL model. 74 extra bytes of JSON took the server's peak RSS from 2 271 MiB to 13 629 MiB over unauthenticated `POST /tokenize`. The `num_frames` ceiling reported in GHSA-vxqj-p4gw-9h4c and fixed by open PR #51969 does not reach this path: the Qwen samplers do not read `num_frames` at all. The same knobs were already capped upstream for `GLMGAVideoBackend` as an accepted security fix in `8b6de0eb9` (PR #54935, merged 2026-09-04); that cap never reached Qwen. ### Details #### Relationship to GHSA-vxqj-p4gw-9h4c and PR #51969 (read this first) GHSA-vxqj-p4gw-9h4c reported that request-level `media_io_kwargs.video.num_frames` overrides the engine frame-count ceiling, and open PR #51969 fixes it by clamping `num_frames` inside `VideoMediaIO.merge_kwargs`. That clamp does not reach the Qwen samplers. `Qwen2VLVideoBackend` and `Qwen3VLVideoBackend` do not read `num_frames` at all — `Qwen2VLVideoBackend`'s own docstring says so ("``num_frames`` is ignored (fps-driven, like the Qwen3-VL loader)"). They bound on `max_frames`, read from the same merged dict with no ceiling: ```python # vllm/multimodal/video.py — Qwen3VLVideoBackend.compute_frames_index_to_sample min_frames = kwargs.get("min_frames", 4) max_frames = kwargs.get("max_frames", 768) num_frames = int(to

Products
vllm
Affected versions
0.24.0, 0.25.0, 0.25.1, 0.26.0, 0.27.0, 0.27.1, 0.28.0, 0.29.0
Fixed versions
0.30.0
CVEs
CVE-2026-105758
Confirmed fixed CVEs
Not specified
Updated in this feed
30 Oct 2026, 18:15 UTC · Revision 1

Source evidence

vLLM: Qwen2-VL / Qwen3-VL video samplers bound on request-controlled max_frames, which the num_frames ceiling does not reach

core-osv · Original source

Published 5 Oct 2026, 23:42 UTC

### Summary An unauthenticated remote attacker can exhaust the memory of the vLLM API-server process by raising the request-level `media_io_kwargs.video.max_frames` and `fps` knobs on any deployment serving a Qwen2-VL or Qwen3-VL model. 74 extra bytes of JSON took the server's peak RSS from 2 271 MiB to 13 629 MiB over unauthenticated `POST /tokenize`. The `num_frames` ceiling reported in GHSA-vxqj-p4gw-9h4c and fixed by open PR #51969 does not reach this path: the Qwen samplers do not read `num_frames` at all. The same knobs were already capped upstream for `GLMGAVideoBackend` as an accepted security fix in `8b6de0eb9` (PR #54935, merged 2026-09-04); that cap never reached Qwen. ### Details #### Relationship to GHSA-vxqj-p4gw-9h4c and PR #51969 (read this first) GHSA-vxqj-p4gw-9h4c reported that request-level `media_io_kwargs.video.num_frames` overrides the engine frame-count ceiling, and open PR #51969 fixes it by clamping `num_frames` inside `VideoMediaIO.merge_kwargs`. That clamp does not reach the Qwen samplers. `Qwen2VLVideoBackend` and `Qwen3VLVideoBackend` do not read `num_frames` at all — `Qwen2VLVideoBackend`'s own docstring says so ("``num_frames`` is ignored (fps-driven, like the Qwen3-VL loader)"). They bound on `max_frames`, read from the same merged dict with no ceiling: ```python # vllm/multimodal/video.py — Qwen3VLVideoBackend.compute_frames_index_to_sample min_frames = kwargs.get("min_frames", 4) max_frames = kwargs.get("max_frames", 768) num_frames = int(to

osv:GHSA-x6mc-67gf-chw4

vllm

Fixed versions: 0.30.0

Complete source record
{
  "id": "25429f2ae087869e14173ca8b16b1a0f7f3b58a5840adefc5b81ad8c7d44b0b3",
  "url": "https://osv.dev/vulnerability/GHSA-x6mc-67gf-chw4",
  "method": "structured",
  "payload": {
    "kind": "advisory",
    "title": "vLLM: Qwen2-VL / Qwen3-VL video samplers bound on request-controlled max_frames, which the num_frames ceiling does not reach",
    "summary": "### Summary An unauthenticated remote attacker can exhaust the memory of the vLLM API-server process by raising the request-level `media_io_kwargs.video.max_frames` and `fps` knobs on any deployment serving a Qwen2-VL or Qwen3-VL model. 74 extra bytes of JSON took the server's peak RSS from 2 271 MiB to 13 629 MiB over unauthenticated `POST /tokenize`. The `num_frames` ceiling reported in GHSA-vxqj-p4gw-9h4c and fixed by open PR #51969 does not reach this path: the Qwen samplers do not read `num_frames` at all. The same knobs were already capped upstream for `GLMGAVideoBackend` as an accepted security fix in `8b6de0eb9` (PR #54935, merged 2026-09-04); that cap never reached Qwen. ### Details #### Relationship to GHSA-vxqj-p4gw-9h4c and PR #51969 (read this first) GHSA-vxqj-p4gw-9h4c reported that request-level `media_io_kwargs.video.num_frames` overrides the engine frame-count ceiling, and open PR #51969 fixes it by clamping `num_frames` inside `VideoMediaIO.merge_kwargs`. That clamp does not reach the Qwen samplers. `Qwen2VLVideoBackend` and `Qwen3VLVideoBackend` do not read `num_frames` at all — `Qwen2VLVideoBackend`'s own docstring says so (\"``num_frames`` is ignored (fps-driven, like the Qwen3-VL loader)\"). They bound on `max_frames`, read from the same merged dict with no ceiling: ```python # vllm/multimodal/video.py — Qwen3VLVideoBackend.compute_frames_index_to_sample min_frames = kwargs.get(\"min_frames\", 4) max_frames = kwargs.get(\"max_frames\", 768) num_frames = int(to",
    "evidence": "osv:GHSA-x6mc-67gf-chw4",
    "products": [
      {
        "name": {
          "value": "vllm",
          "evidence": "/affected/0/package/name"
        },
        "prerequisites": [],
        "fixed_versions": [
          {
            "value": "0.30.0",
            "evidence": "/affected/0/ranges"
          }
        ],
        "affected_versions": [
          {
            "value": "0.24.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.25.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.25.1",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.26.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.27.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.27.1",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.28.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.29.0",
            "evidence": "/affected/0/versions"
          }
        ]
      }
    ],
    "severity": "medium",
    "structured": {
      "id": "GHSA-x6mc-67gf-chw4",
      "aliases": [
        "CVE-2026-105758",
        "GHSA-x6mc-67gf-chw4",
        "PYSEC-2026-4310"
      ],
      "metrics": [
        {
          "type": "CVSS_V3",
          "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L"
        }
      ],
      "affected": [
        {
          "ranges": [
            {
              "type": "ECOSYSTEM",
              "events": [
                {
                  "introduced": "0.24.0"
                },
                {
                  "fixed": "0.30.0"
                }
              ]
            }
          ],
          "package": {
            "name": "vllm",
            "purl": "pkg:pypi/vllm",
            "ecosystem": "PyPI"
          },
          "versions": [
            "0.24.0",
            "0.25.0",
            "0.25.1",
            "0.26.0",
            "0.27.0",
            "0.27.1",
            "0.28.0",
            "0.29.0"
          ],
          "database_specific": {
            "source": "https://github.com/github/advisory-database/blob/main/advisories/github-reviewed/2026/10/GHSA-x6mc-67gf-chw4/GHSA-x6mc-67gf-chw4.json"
          }
        }
      ],
      "modified": "2026-10-15T14:55:52.413199549Z",
      "provider": "osv",
      "references": [
        "https://github.com/vllm-project/vllm/security/advisories/GHSA-x6mc-67gf-chw4",
        "https://github.com/vllm-project/vllm/pull/56729",
        "https://github.com/vllm-project/vllm/commit/ea723c81c3ea26425cb69503a5d5e90822a04a45",
        "https://github.com/vllm-project/vllm",
        "https://github.com/vllm-project/vllm/releases/tag/v0.30.0"
      ],
      "provider_metadata": {
        "cwe_ids": [
          "CWE-770"
        ],
        "severity": "MODERATE",
        "github_reviewed": true,
        "nvd_published_at": null,
        "github_reviewed_at": "2026-10-05T23:42:59Z"
      }
    },
    "related_cves": [
      {
        "value": "CVE-2026-105758",
        "evidence": "/aliases"
      }
    ],
    "security_relevant": true,
    "severity_evidence": "/severity/0/score"
  },
  "source_id": "core-osv",
  "document_id": 332682,
  "published_at": "2026-10-05T23:42:59+00:00"
}

vLLM: Qwen2-VL / Qwen3-VL video samplers bound on request-controlled max_frames, which the num_frames ceiling does not reach

core-osv · Original source

Published 15 Oct 2026, 10:22 UTC

### Summary An unauthenticated remote attacker can exhaust the memory of the vLLM API-server process by raising the request-level `media_io_kwargs.video.max_frames` and `fps` knobs on any deployment serving a Qwen2-VL or Qwen3-VL model. 74 extra bytes of JSON took the server's peak RSS from 2 271 MiB to 13 629 MiB over unauthenticated `POST /tokenize`. The `num_frames` ceiling reported in GHSA-vxqj-p4gw-9h4c and fixed by open PR #51969 does not reach this path: the Qwen samplers do not read `num_frames` at all. The same knobs were already capped upstream for `GLMGAVideoBackend` as an accepted security fix in `8b6de0eb9` (PR #54935, merged 2026-09-04); that cap never reached Qwen. ### Details #### Relationship to GHSA-vxqj-p4gw-9h4c and PR #51969 (read this first) GHSA-vxqj-p4gw-9h4c reported that request-level `media_io_kwargs.video.num_frames` overrides the engine frame-count ceiling, and open PR #51969 fixes it by clamping `num_frames` inside `VideoMediaIO.merge_kwargs`. That clamp does not reach the Qwen samplers. `Qwen2VLVideoBackend` and `Qwen3VLVideoBackend` do not read `num_frames` at all — `Qwen2VLVideoBackend`'s own docstring says so ("``num_frames`` is ignored (fps-driven, like the Qwen3-VL loader)"). They bound on `max_frames`, read from the same merged dict with no ceiling: ```python # vllm/multimodal/video.py — Qwen3VLVideoBackend.compute_frames_index_to_sample min_frames = kwargs.get("min_frames", 4) max_frames = kwargs.get("max_frames", 768) num_frames = int(to

osv:PYSEC-2026-4310

vllm

Fixed versions: 0.30.0

Complete source record
{
  "id": "5d86939e9ce3ea1b205982bd431579250dabd2fcff7234110222b1134fe21493",
  "url": "https://osv.dev/vulnerability/PYSEC-2026-4310",
  "method": "structured",
  "payload": {
    "kind": "advisory",
    "title": "vLLM: Qwen2-VL / Qwen3-VL video samplers bound on request-controlled max_frames, which the num_frames ceiling does not reach",
    "summary": "### Summary An unauthenticated remote attacker can exhaust the memory of the vLLM API-server process by raising the request-level `media_io_kwargs.video.max_frames` and `fps` knobs on any deployment serving a Qwen2-VL or Qwen3-VL model. 74 extra bytes of JSON took the server's peak RSS from 2 271 MiB to 13 629 MiB over unauthenticated `POST /tokenize`. The `num_frames` ceiling reported in GHSA-vxqj-p4gw-9h4c and fixed by open PR #51969 does not reach this path: the Qwen samplers do not read `num_frames` at all. The same knobs were already capped upstream for `GLMGAVideoBackend` as an accepted security fix in `8b6de0eb9` (PR #54935, merged 2026-09-04); that cap never reached Qwen. ### Details #### Relationship to GHSA-vxqj-p4gw-9h4c and PR #51969 (read this first) GHSA-vxqj-p4gw-9h4c reported that request-level `media_io_kwargs.video.num_frames` overrides the engine frame-count ceiling, and open PR #51969 fixes it by clamping `num_frames` inside `VideoMediaIO.merge_kwargs`. That clamp does not reach the Qwen samplers. `Qwen2VLVideoBackend` and `Qwen3VLVideoBackend` do not read `num_frames` at all — `Qwen2VLVideoBackend`'s own docstring says so (\"``num_frames`` is ignored (fps-driven, like the Qwen3-VL loader)\"). They bound on `max_frames`, read from the same merged dict with no ceiling: ```python # vllm/multimodal/video.py — Qwen3VLVideoBackend.compute_frames_index_to_sample min_frames = kwargs.get(\"min_frames\", 4) max_frames = kwargs.get(\"max_frames\", 768) num_frames = int(to",
    "evidence": "osv:PYSEC-2026-4310",
    "products": [
      {
        "name": {
          "value": "vllm",
          "evidence": "/affected/0/package/name"
        },
        "prerequisites": [],
        "fixed_versions": [
          {
            "value": "0.30.0",
            "evidence": "/affected/0/ranges"
          }
        ],
        "affected_versions": [
          {
            "value": "0.24.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.25.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.25.1",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.26.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.27.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.27.1",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.28.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "0.29.0",
            "evidence": "/affected/0/versions"
          }
        ]
      }
    ],
    "severity": "medium",
    "structured": {
      "id": "PYSEC-2026-4310",
      "aliases": [
        "CVE-2026-105758",
        "GHSA-x6mc-67gf-chw4",
        "PYSEC-2026-4310"
      ],
      "metrics": [
        {
          "type": "CVSS_V3",
          "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L"
        }
      ],
      "affected": [
        {
          "ranges": [
            {
              "type": "ECOSYSTEM",
              "events": [
                {
                  "introduced": "0.24.0"
                },
                {
                  "fixed": "0.30.0"
                }
              ]
            }
          ],
          "package": {
            "name": "vllm",
            "purl": "pkg:pypi/vllm",
            "ecosystem": "PyPI"
          },
          "versions": [
            "0.24.0",
            "0.25.0",
            "0.25.1",
            "0.26.0",
            "0.27.0",
            "0.27.1",
            "0.28.0",
            "0.29.0"
          ],
          "database_specific": {
            "source": "https://github.com/pypa/advisory-database/blob/main/vulns/vllm/PYSEC-2026-4310.yaml"
          }
        }
      ],
      "modified": "2026-10-15T14:45:14.128616170Z",
      "provider": "osv",
      "references": [
        "https://github.com/vllm-project/vllm/security/advisories/GHSA-x6mc-67gf-chw4",
        "https://github.com/vllm-project/vllm/pull/56729",
        "https://github.com/vllm-project/vllm/commit/ea723c81c3ea26425cb69503a5d5e90822a04a45",
        "https://github.com/vllm-project/vllm",
        "https://github.com/vllm-project/vllm/releases/tag/v0.30.0",
        "https://pypi.org/project/vllm",
        "https://github.com/advisories/GHSA-x6mc-67gf-chw4",
        "https://nvd.nist.gov/vuln/detail/CVE-2026-105758"
      ]
    },
    "related_cves": [
      {
        "value": "CVE-2026-105758",
        "evidence": "/aliases"
      }
    ],
    "security_relevant": true,
    "severity_evidence": "/severity/0/score"
  },
  "source_id": "core-osv",
  "document_id": 332658,
  "published_at": "2026-10-15T10:22:45.585356+00:00"
}