CVE-2026-55514CWE-617

vLLM denial of service via prompt embeds on M-RoPE models

High · published July 6, 2026

CVSS v4.0
7.1
EPSS
1%
Percentile
51.2
In the wild
Unconfirmed
What it is

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.

The record
Technical detail
CVSS v4.0
7.1 · HIGH
Vector
CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N
EPSS
0.00665 · 51.2th percentile
Weakness
CWE-617 · Reachable Assertion
Published
2026-07-06T20:07Z
EPSS history
Timeline
  • 06 JUL 20:07Z
    vLLM denial of service via prompt embeds on M-RoPE models
    cvelistv5