Lucene search
+L

CVE-2026-100651

🗓️ 26 Sep 2026 13:23:21Reported by VulnCheckType 
cve
 cve
🔗 web.nvd.nist.gov👁 33 Views🌐 WEB

vLLM before 0.29.0 DoS in disaggregated serving via overlong multimodal token_ids.

Related
Detection
Affected
Refs
Paths
ReporterTitlePublishedViews
Family
cvelist
Cvelist
CVE-2026-100651 vllm before 0.29.0 Denial of Service via Decoder Prompt Length Bypass
26 Sep 202613:23
–cvelist
euvd
EUVD
EUVD-2026-87745
26 Sep 202613:23
–euvd
mend
Mend
CVE-2026-100651
26 Sep 202613:23
–mend
nvd
NVD
CVE-2026-100651
26 Sep 202614:16
–nvd
osv_mini
osv_mini
MINI-2V6J-5Q34-3J32
1 Oct 202614:12
–osv_mini
osv_mini
osv_mini
MINI-472M-WVCF-F92X
1 Oct 202614:12
–osv_mini
osv_mini
osv_mini
MINI-XG3F-RF8R-FR69
1 Oct 202615:25
–osv_mini
ptsecurity
Positive Technologies
PT-2026-99322
26 Sep 202600:00
–ptsecurity
rapid7
Rapid7 Vulnerability Database (full)
CVE-2026-100651: Uncontrolled Resource Consumption
26 Sep 202600:00
–rapid7
redhatcve
RedhatCVE
CVE-2026-100651
26 Sep 202613:48
–redhatcve
Rows per page
NVD
Vulners
CNA
Node
vllmvllmRange<0.29.0
[
  {
    "vendor": "vllm-project",
    "product": "vllm",
    "defaultStatus": "unaffected",
    "packageURL": "pkg:pypi/vllm",
    "versions": [
      {
        "version": "0",
        "status": "affected",
        "versionType": "semver",
        "lessThan": "0.29.0"
      },
      {
        "version": "0.29.0",
        "status": "unaffected",
        "versionType": "semver"
      }
    ]
  }
]
ParameterPositionPathDescriptionCWE
featuresrequest body/inference/v1/generatevLLM before 0.29.0 fails to enforce decoder prompt-length validation on the disaggregated serving endpoint /inference/v1/generate when a multimodal payload ('features') is provided; token_ids are not checked against model_config.max_model_len, allowing an overlong prompt to reach the worker and potentially cause DoS. Fixed in 0.29.0.CWE-400
token_idsrequest body/inference/v1/generatevLLM before 0.29.0 fails to enforce decoder prompt-length validation on the disaggregated serving endpoint /inference/v1/generate when a multimodal payload ('features') is provided; token_ids are not checked against model_config.max_model_len, allowing an overlong prompt to reach the worker and potentially cause DoS. Fixed in 0.29.0.CWE-400

Data

Build on a solid foundation with Vulners data

We provide the essential building blocks for cybersecurity solutions with comprehensive, structured, and constantly updated vulnerability and exploits data

Api

Power your application with Vulners API

The Vulners REST API offers reliable, high-performance access to vulnerability intelligence, with 99.9% SLA uptime and CDN-backed data delivery for seamless global access

App

Assess and manage vulnerabilities with Vulners tools

Built on top of Vulners' database and SDK, end-user solutions give security professionals and developers lightweight and powerful tools for vulnerability remediation

06 Oct 2026 19:30Current
5.8Medium risk
Vulners AI Score5.8
CVSS 3.16.5
CVSS 47.1
EPSS0.00331
SSVC
33