CVE-2026-115916

vLLM: Out-of-vocabulary prompt token id passes input validation and triggers a CUDA device-side assert (engine crash / DoS)

Published Oct 26, 2026 CVSS 3.1 · 6.5 Medium EPSS < 1% KEV · Not listed
Description
vLLM is an inference and serving engine for large language models. Prior to 0.32.0, vllm/tokenizers/hf.py computes max_token_id with an off-by-one upper bound, allowing a pre-tokenized prompt ID equal to the vocabulary size to pass validation. For a tensor-parallel size of one and an exact multiple-of-64 vocabulary, VocabParallelEmbedding.forward does not mask that value before F.embedding uses it, causing a CUDA device-side assertion and terminating the engine. This issue is fixed in version 0.32.0.

Analysis

AI-generated

Technical impact

vLLM calculates the maximum token identifier with an off‑by‑one error, allowing a pre‑tokenized prompt that has an ID equal to the vocabulary size to be considered valid. When the inference engine processes such a prompt in a single‑rank, tensor‑parallel setting with an exact multiple‑of‑64 vocabulary, the embedding layer receives an out‑of‑range ID that triggers a CUDA device‑side assertion, causing the engine to crash. The result is an availability failure that can be repeated by submitting a crafted prompt, leading to a denial‑of‑service condition for the service.

Affected Systems

The vulnerability exists in all vLLM releases prior to version 0.32.0. Any deployment that uses the affected tokenization logic and the tensor‑parallel embedding path with a vocab size that is a multiple of 64 is impacted. The product is vLLM, an inference and serving engine for large language models.

Risk and Exploitability

The CVSS score of 6.5 indicates a moderate severity for the availability impact. The EPSS score is not available, so the precise likelihood of exploitation is unknown, and the vulnerability is not listed in the CISA KEV catalog. The probable attack vector is the submission of a malicious prompt to an inference endpoint that accepts untrusted input, as the off‑by‑one validation flaw can be triggered from the network. No other access or privilege escalation is described, but repeated crashes can disrupt service availability.

Remediation

No solution or workaround provided in the CVE record.

OpenCVE Recommended Actions

  • Upgrade vLLM to version 0.32.0 or later, where the off‑by‑one error is fixed.
  • If an upgrade is not possible, intercept incoming prompts and validate token identifiers, rejecting any that are equal to or greater than the vocabulary size before they reach the embedding layer.
  • Apply deployment hardening measures such as limiting the request size, rate‑limiting clients, or disabling tensor‑parallel embedding for single‑rank runs as a temporary workaround until the patch can be applied.

Generated by OpenCVE AI on October 26, 2026 at 18:23 UTC.

Tracking

Sign in to view the affected projects.

Advisories

Source ID Title
Github GHSA Github GHSA GHSA-q97c-p368-75c4 vLLM: Out-of-vocabulary prompt token id passes input validation and triggers a CUDA device-side assert (engine crash / DoS)
History
Type Values Removed Values Added
Metrics ssvc

{'options': {'Automatable': 'no', 'Exploitation': 'none', 'Technical Impact': 'partial'}, 'version': '2.0.3'}

Type Values Removed Values Added
CPEs cpe:2.3:a:vllm:vllm:*:*:*:*:*:*:*:*
Type Values Removed Values Added
First Time appeared Vllm
Vllm vllm
Vllm-project
Vllm-project vllm
Vendors & Products Vllm
Vllm vllm
Vllm-project
Vllm-project vllm
Type Values Removed Values Added
Description vLLM is an inference and serving engine for large language models. Prior to 0.32.0, vllm/tokenizers/hf.py computes max_token_id with an off-by-one upper bound, allowing a pre-tokenized prompt ID equal to the vocabulary size to pass validation. For a tensor-parallel size of one and an exact multiple-of-64 vocabulary, VocabParallelEmbedding.forward does not mask that value before F.embedding uses it, causing a CUDA device-side assertion and terminating the engine. This issue is fixed in version 0.32.0.
Title vLLM: Out-of-vocabulary prompt token id passes input validation and triggers a CUDA device-side assert (engine crash / DoS)
Weaknesses CWE-129
CWE-193
References
Metrics cvssV3_1

{'score': 6.5, 'vector': 'CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H'}

Weaknesses

Sources

Status
PUBLISHED
Assigner
GitHub_M
Published
Updated
Nov 2, 2026, 5:50 PM
Reserved
Oct 23, 2026, 8:21 PM
Status
Analyzed
Published
Oct 26, 2026, 6:17 PM
Modified
Nov 2, 2026, 6:18 PM
Severity
Moderate
Public date
Oct 26, 2026, 5:18 PM