---
id: CVE-2026-105754
title: vLLM is an inference and serving engine for large language models
summary: >-
  vLLM is an inference and serving engine for large language models. Prior to
  0.30.0, the /inference/v1/generate endpoint in the disaggregated scale-out
  path accepts caller-supplied tensors in the features.kwargs_data field, cache
  identifi…
severity: medium
cvss: 6.5
cvssVector: 'CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H'
cwe:
  - CWE-20
  - CWE-617
  - CWE-639
  - CWE-668
  - CWE-704
  - CWE-1284
vendor: vllm
product: vllm
affected:
  - vllm < 0.30.0
patched:
  - vllm 0.30.0
published: '2026-10-05'
updated: '2026-10-08'
sourceUpdated: '2026-10-08T03:16:35.230'
source: NVD
sourceUrl: 'https://nvd.nist.gov/vuln/detail/CVE-2026-105754'
references:
  - url: >-
      https://github.com/vllm-project/vllm/commit/1970f3ed4be7fa8620e4ddc4a12c36a8384cfc27
    label: security-advisories@github.com
  - url: 'https://github.com/vllm-project/vllm/pull/51898'
    label: security-advisories@github.com
  - url: 'https://github.com/vllm-project/vllm/releases/tag/v0.30.0'
    label: security-advisories@github.com
  - url: >-
      https://github.com/vllm-project/vllm/security/advisories/GHSA-ph72-cqr5-qpp7
    label: security-advisories@github.com
  - url: 'https://github.com/advisories/GHSA-ph72-cqr5-qpp7'
  - url: 'https://github.com/vllm-project/vllm'
  - url: >-
      https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-105754.json
  - url: 'https://access.redhat.com/security/cve/CVE-2026-105754'
  - url: 'https://bugzilla.redhat.com/show_bug.cgi?id=2546186'
  - url: 'https://www.cve.org/CVERecord?id=CVE-2026-105754'
  - url: 'https://nvd.nist.gov/vuln/detail/CVE-2026-105754'
  - url: 'https://pypi.org/project/vllm'
tags:
  - nvd
  - cve.org
  - ghsa
  - pip
  - osv
  - csaf
  - vex
  - red-hat
epss: 0.00269
epssPercentile: 0.18966
aliases:
  - GHSA-ph72-cqr5-qpp7
  - PYSEC-2026-4309
ecosystem: pip
ssvc:
  exploitation: none
  automatable: 'no'
  technicalImpact: partial
  timestamp: '2026-10-08T02:31:20.281715Z'
ingestedAt: '2026-10-05T23:36:21.156Z'
---

## Overview

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the /inference/v1/generate endpoint in the disaggregated scale-out path accepts caller-supplied tensors in the features.kwargs_data field, cache identifiers in the features.mm_hashes field, ranges in the features.mm_placeholders field, and wire-selected multimodal field processors without rebinding them to the active model renderer contract. Forged grid geometry, field types, or non-positive placeholder lengths can terminate the shared EngineCore; when an attacker knows or can induce a victim's content hash, forged cache hashes can poison or retrieve cross-request encoder-cache state; and dropped sparse placeholder masks can alter replayed transport semantics. This issue is fixed in version 0.30.0.

## Affected

- `vllm < 0.30.0`

## Remediation

Upgrade past the affected range:

- `vllm 0.30.0`

## Package advisory (CVE-2026-105754)

Affected packages:

- `vllm < 0.30.0`

Patched in:

- `vllm 0.30.0`

Source: https://github.com/advisories/GHSA-ph72-cqr5-qpp7

## Vendor advisories

- **Red Hat VEX** · Moderate · affected: Red Hat AI Inference Server, Red Hat Enterprise Linux AI (RHEL AI) 3, Red Hat OpenShift AI (RHOAI) · no fix planned: Red Hat AI Inference Server, Red Hat Enterprise Linux AI (RHEL AI) 3, Red Hat OpenShift AI (RHOAI) · updated 2026-10-06 · [vex](https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-105754.json)