Skip to content
Heist/ feed
Back to feed
Medium

vLLM: Incomplete CVE-2025-62164 remediation can be bypassed by concurrent prompt parts

Type: advisorySources: OSV

vLLM is an inference and serving engine for large language models. From 0.20.2rc0 until 0.26.0, safe_load_prompt_embeds in vllm/renderers/embed_utils.py uses torch.sparse.check_sparse_tensor_invariants, whose process-global save, enable, and restore state can be raced by concurrent prompt_embeds parts submitted to POST /v1/chat/completions through AsyncMultiModalItemTracker.resolve_items, asyncio.gather, and the default executor, allowing an invalid sparse tensor to reach tensor.to_dense despite the CVE-2025-62164 guard when enable_prompt_embeds is enabled. This issue is fixed in version 0.26.0.

Products
Not specified
Affected versions
Not specified
Fixed versions
Not specified
CVEs
CVE-2026-73557
Confirmed fixed CVEs
Not specified
Updated in this feed
3 Nov 2026, 08:15 UTC · Revision 1

Source evidence

vLLM: Incomplete CVE-2025-62164 remediation can be bypassed by concurrent prompt parts

core-osv · Original source

Published 13 Aug 2026, 15:00 UTC

vLLM is an inference and serving engine for large language models. From 0.20.2rc0 until 0.26.0, safe_load_prompt_embeds in vllm/renderers/embed_utils.py uses torch.sparse.check_sparse_tensor_invariants, whose process-global save, enable, and restore state can be raced by concurrent prompt_embeds parts submitted to POST /v1/chat/completions through AsyncMultiModalItemTracker.resolve_items, asyncio.gather, and the default executor, allowing an invalid sparse tensor to reach tensor.to_dense despite the CVE-2025-62164 guard when enable_prompt_embeds is enabled. This issue is fixed in version 0.26.0.

osv:CVE-2026-73557
Complete source record
{
  "id": "29be51b64e227a31a4cfc5bd6c075cc94f2a121628ed74869a7259ac0ddf329c",
  "url": "https://osv.dev/vulnerability/CVE-2026-73557",
  "method": "structured",
  "payload": {
    "kind": "advisory",
    "title": "vLLM: Incomplete CVE-2025-62164 remediation can be bypassed by concurrent prompt parts",
    "summary": "vLLM is an inference and serving engine for large language models. From 0.20.2rc0 until 0.26.0, safe_load_prompt_embeds in vllm/renderers/embed_utils.py uses torch.sparse.check_sparse_tensor_invariants, whose process-global save, enable, and restore state can be raced by concurrent prompt_embeds parts submitted to POST /v1/chat/completions through AsyncMultiModalItemTracker.resolve_items, asyncio.gather, and the default executor, allowing an invalid sparse tensor to reach tensor.to_dense despite the CVE-2025-62164 guard when enable_prompt_embeds is enabled. This issue is fixed in version 0.26.0.",
    "evidence": "osv:CVE-2026-73557",
    "products": [],
    "severity": "medium",
    "structured": {
      "id": "CVE-2026-73557",
      "aliases": [
        "CVE-2026-73557",
        "GHSA-pr7f-p5mw-fc87",
        "PYSEC-2026-3938"
      ],
      "metrics": [
        {
          "type": "CVSS_V4",
          "score": "CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N"
        }
      ],
      "affected": [
        {
          "ranges": [
            {
              "repo": "https://github.com/vllm-project/vllm",
              "type": "GIT",
              "events": [
                {
                  "introduced": "9da56fd18b95ff1b11360e8400f7c41b126d190b"
                },
                {
                  "fixed": "ffd46bfab2128bb84146050e98b51a617c6575ab"
                },
                {
                  "fixed": "793cf79c89d4049124e756915468ac30318f2e50"
                }
              ],
              "database_specific": {
                "extracted_events": [
                  {
                    "cpe": "cpe:2.3:a:vllm:vllm:*:*:*:*:*:*:*:*",
                    "range": [
                      {
                        "introduced": "0.21.0"
                      },
                      {
                        "fixed": "0.26.0"
                      }
                    ],
                    "source": "CPE_RANGE"
                  },
                  {
                    "range": [
                      {
                        "fixed": "793cf79c89d4049124e756915468ac30318f2e50"
                      }
                    ],
                    "source": "REFERENCES_COMMIT"
                  },
                  {
                    "range": [
                      {
                        "fixed": "ffd46bfab2128bb84146050e98b51a617c6575ab"
                      }
                    ],
                    "source": "REFERENCES_TAG",
                    "original_tag": "v0.26.0"
                  }
                ]
              }
            }
          ],
          "database_specific": {
            "source": "https://storage.googleapis.com/cve-osv-conversion/osv-output/CVE-2026-73557.json"
          }
        }
      ],
      "modified": "2026-10-28T02:30:27.279721406Z",
      "provider": "osv",
      "references": [
        "https://github.com/vllm-project/vllm/releases/tag/v0.26.0",
        "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/73xxx/CVE-2026-73557.json",
        "https://github.com/vllm-project/vllm/security/advisories/GHSA-pr7f-p5mw-fc87",
        "https://nvd.nist.gov/vuln/detail/CVE-2026-73557",
        "https://github.com/vllm-project/vllm/commit/793cf79c89d4049124e756915468ac30318f2e50",
        "https://github.com/vllm-project/vllm/pull/48583"
      ],
      "provider_metadata": {
        "cwe_ids": [
          "CWE-362"
        ],
        "cna_assigner": "GitHub_M",
        "osv_generated_from": "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/73xxx/CVE-2026-73557.json"
      }
    },
    "related_cves": [
      {
        "value": "CVE-2026-73557",
        "evidence": "/id"
      }
    ],
    "security_relevant": true,
    "severity_evidence": "/severity/0/score"
  },
  "source_id": "core-osv",
  "document_id": 493701,
  "published_at": "2026-08-13T15:00:15.746+00:00"
}