CVE-2026-55514

vllm-project vllmVulnerabilityCVSS 7.1

vLLM is a library for LLM inference and serving. From 0.12.0 to before 0.24.0, sending a pure prompt embeds payload in a /v1/completions request with a model using M-RoPE causes EngineCore to fail an assertion and fatally crash, shutting down the entire server application. Any remote user who is authorized to make a /v1/completions request can make such a request and induce a crash. This issue is fixed in version 0.24.0.

No public exploitation has been reported so far. No distribution fix is listed yet; apply vendor mitigations and monitor.

Recommended action: Track your distribution's security tracker and apply the fix as soon as it ships.

CVE-2026-55514 vllm-project vllm Vulnerability

Severity metrics

CVSS 4.0 : 7.1 HIGH · CVSS 3.1 : 6.5 MEDIUM

Attack vectorNetwork (remote)

Attack complexityLow

Attack requirementsNone

Privileges requiredLow

User interactionNone

Confidentiality impactNone

Integrity impactNone

Availability impactHigh

Subsequent confidentialityNone

Subsequent integrityNone

Subsequent availabilityNone

Vector: AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N

Affected & fixed versions by distribution

Versions are the native distribution strings (backport-aware): RHEL/Debian/SUSE backport fixes, so compare with the distro version comparator (dpkg / rpm / vercmp), not the upstream version.

No affected package in the tracked watchlist yet (Debian, Ubuntu, Arch, SUSE, Red Hat...). Triage may still be pending.

Sources

Advisory NVD CVE.org

Published 2026-07-06 · updated 2026-07-27 · source: NVD