Read the docs
TI-2026-80517 Aliases CVE-2026-115916GHSA-q97c-p368-75c4

Out-of-vocabulary prompt token IDs crash vLLM engine

Priority LOW 29CWE-129Published · Updated

Key facts

Attack vector
Network
Privileges required
Low
User interaction
None
Attack complexity
Low
Impact type
Denial of service
Confidentiality · integrity · availability
None · None · High
Fix status
Partial Red Hat: 3 without a fix

Attack profile from CVSS 3.1 · CNA. Fix status from CNA, GitHub and Red Hat. Impact type from CWE-129 and the CVSS impact.

Summary

vLLM versions before 0.32.0 are affected. An off-by-one error in the maximum token ID check (CWE-193) lets a prompt token ID equal to the vocabulary size pass validation. With tensor-parallel size one and a vocabulary size that is an exact multiple of 64, the embedding layer does not mask that value, which triggers a CUDA device-side assertion and terminates the engine. A remote user with low privileges can crash the inference engine.

Written with Claude Haiku 5.5 from the sources below · 2026-11-02

Description from the CNA

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.

Why it matters now

Exploitation None reported CISA SSVC · not in CISA KEV
EPSS 0.00312 24th percentileScored 2026-11-07
Fix Partial vllm (vllm-project): 0.32.0vllm (PyPI): 0.32.0Red Hat: 3 without a fix

SSVC decision points

CISA ADP · SSVC 2.0.3

Exploitation
None
Automatable
No
Technical impact
Partial

The decision (Track, Attend, Act) also depends on your mission and well-being impact; see CISA's SSVC guide.

Affected products

VendorProductAffected versionsFixed inStatusSource
vllm-projectvllmbefore 0.32.00.32.0Patch availableCNA
—vllm (PyPI)before 0.32.00.32.0Patch availableGHSA-q97c-p368-75c4
  • vllm Vendorvllm-project Affected before 0.32.0 Fixed in 0.32.0 Patch available CNA
  • vllm (PyPI) Vendor— Affected before 0.32.0 Fixed in 0.32.0 Patch available GHSA-q97c-p368-75c4

CPE configurations NVD · 1

CPEVersionsRoleVendor
cpe:2.3:a:vllm:vllm:*:*:*:*:*:*:*:*before 0.32.0VulnerablevLLM
  • cpe:2.3:a:vllm:vllm:*:*:*:*:*:*:*:* Versions before 0.32.0 Vulnerable ·vLLM

Remediation

Upgrade vLLM to version 0.32.0 on PyPI. Red Hat AI Inference Server, Red Hat Enterprise Linux AI (RHEL AI) 3 and Red Hat OpenShift AI (RHOAI) are affected, and Red Hat gives no plan for them.

Written with Claude Haiku 5.5 from the sources below · 2026-11-02

  • Fix · CNA vllm (vllm-project) Fixed in 0.32.0
  • Fix · GHSA-q97c-p368-75c4 vllm (PyPI) Fixed in 0.32.0 Advisory →

Vendor assessments

Red Hat

Red Hat rating: Moderate. Red Hat's page for this CVE →

ProductStatusPlan or fix
Red Hat AI Inference ServerAffectedNo fix yet: Fix deferred
Red Hat Enterprise Linux AI (RHEL AI) 3AffectedNo fix yet: Fix deferred
Red Hat OpenShift AI (RHOAI)AffectedNo fix yet: Fix deferred

Timeline

  1. 2026-10-26 Threatist record created from CVE-2026-115916
  2. 2026-10-27 Affected versions changed · Red Hat
  3. 2026-10-27 CVSS score added · Red Hat
  4. 2026-10-27 Priority changed: MEDIUM → LOW
  5. 2026-10-28 Alias added: GHSA-q97c-p368-75c4
  6. 2026-10-28 Fix available: PyPI/[email protected] · GHSA-q97c-p368-75c4
  7. 2026-10-28 CVSS score added · GHSA-q97c-p368-75c4
  8. 2026-10-28 Weakness changed · GHSA-q97c-p368-75c4
  9. 2026-10-28 6 new references · GHSA-q97c-p368-75c4
  10. 2026-11-02 CISA SSVC decision points changed

References

Patch 3

Advisory 1

Other 2

  • TI-2024-10229 Remote Code Execution in vllm-project/vllm CVSS 3.0CRITICAL 9.8
  • TI-2024-08404 Remote Code Execution in vllm-project/vllm CVSS 3.1CRITICAL 9.8
  • TI-2026-36017 vLLM: OpenAI auth bypass CVSS 3.1CRITICAL 9.1
  • TI-2026-63056 vLLM before 0.28.0 Denial of Service via negative token ID CVSS 4.0HIGH 8.7
  • TI-2026-04308 Hardcoded trust_remote_code=True in vllm-project/vllm Bypasses User Security Control CVSS 3.0HIGH 8.8
  • TI-2026-39137 vLLM: Remote DoS in vLLM via Invalid Recovered Token Reinjection CVSS 3.1HIGH 7.5
See all 124 →

Sources

This product uses the NVD API but is not endorsed or certified by the NVD. Sources and licences