---
id: CVE-2026-100651
title: >-
  vLLM before 0.29.0 fails to enforce decoder prompt-length validation on the
  disaggregated serving endpoint /inference/v1/generate
summary: >-
  vLLM before 0.29.0 fails to enforce decoder prompt-length validation on the
  disaggregated serving endpoint /inference/v1/generate. When the request
  contains a 'features' (multimodal) payload,
  vllm/entrypoints/serve/disagg/serving.py buil…
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-400
  - CWE-120
vendor: vllm
product: vllm
affected:
  - vllm < 0.29.0
patched:
  - vllm 0.29.0
published: '2026-09-26'
updated: '2026-10-06'
sourceUpdated: '2026-10-06T19:30:25.503'
source: NVD
sourceUrl: 'https://nvd.nist.gov/vuln/detail/CVE-2026-100651'
references:
  - url: >-
      https://github.com/vllm-project/vllm/security/advisories/GHSA-3mqx-f33v-vgp9
    label: disclosure@vulncheck.com
  - url: >-
      https://www.vulncheck.com/advisories/vllm-before-0.29.0-denial-of-service-via-decoder-prompt-length-bypass
    label: disclosure@vulncheck.com
  - url: >-
      https://github.com/vllm-project/vllm/security/advisories/GHSA-3mqx-f33v-vgp9
    label: 134c704f-9b21-4f2e-91b3-4a467353bcc0
  - url: >-
      https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-100651.json
  - url: 'https://access.redhat.com/security/cve/CVE-2026-100651'
  - url: 'https://bugzilla.redhat.com/show_bug.cgi?id=2541835'
  - url: 'https://www.cve.org/CVERecord?id=CVE-2026-100651'
  - url: 'https://nvd.nist.gov/vuln/detail/CVE-2026-100651'
tags:
  - nvd
  - cve.org
  - exploit-available
  - csaf
  - vex
  - red-hat
epss: 0.00384
epssPercentile: 0.3231
exploitAvailable: true
ssvc:
  exploitation: poc
  automatable: 'no'
  technicalImpact: partial
  timestamp: '2026-09-28T18:21:26.899493Z'
ingestedAt: '2026-09-26T13:33:20.919Z'
---

## Overview

vLLM before 0.29.0 fails to enforce decoder prompt-length validation on the disaggregated serving endpoint /inference/v1/generate. When the request contains a 'features' (multimodal) payload, vllm/entrypoints/serve/disagg/serving.py builds a multimodal EngineInput directly from the caller-supplied token_ids, and GenerateRequest.token_ids (vllm/entrypoints/serve/disagg/protocol.py) is not checked against model_config.max_model_len. For multimodal processors that report skip_prompt_length_check=True (for example Nemotron Parse, Whisper, and FireRedLID), InputProcessor._validate_prompt_len() returns immediately for both encoder and decoder prompts, so an overlong prompt becomes an EngineCoreRequest and reaches the worker input-batch copy into a fixed max_model_len-wide NumPy row. A client able to reach the endpoint on an affected model configuration can therefore submit an overlong token_ids list to trigger a worker failure and denial of service. Fixed in 0.29.0.

## Affected

- `vllm < 0.29.0`

## Remediation

Upgrade past the affected range:

- `vllm 0.29.0`

## 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 2026-09-27 · [vex](https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-100651.json)