High · CVSS 8.7CVE-2026-56340GHSA-MCMC-2M55-J8JJ

vLLM - Denial of Service via Unvalidated Multimodal EmbeddingsCVE-2026-56340

Answer in brief

CVE-2026-56340 records a High severity (CVSS 8.7) vulnerability in vLLM - Denial of Service via Unvalidated Multimodal Embeddings. The current sources do not mark it as known exploited. The current feed maps Red Hat/rhaiis/vllm-neuron-rhel9 (generic), Red Hat/rhaiis/vllm-spyre-rhel9 (generic), Red Hat/rhaiis/vllm-tpu-rhel9 (generic), vLLM/vLLM (generic) and additional mapped packages. Check affected ranges and fixed versions before updating.

Analysis pending evidence review

HOL Guard separates source facts from reviewed analysis. See the methodology.

Published Jun 20, 2026Updated Nov 8, 2026Source checked Nov 8, 2026First seen by HOL Jun 20, 2026Material review Nov 8, 2026

Upstream Advisory

Affected scope and exposure questions

The current feed maps Red Hat/rhaiis/vllm-neuron-rhel9 (generic), Red Hat/rhaiis/vllm-spyre-rhel9 (generic), Red Hat/rhaiis/vllm-tpu-rhel9 (generic), vLLM/vLLM (generic) and additional mapped packages. Check affected ranges and fixed versions before updating.

Affected products and reported versions
Product	Affected versions	Fixed versions
cpe:2.3:a:vllm:vllm::::::::	Not reported	Not reported

Mapped affected packages and fixed versions
Package	Affected range	Fixed version
Red Hat/rhaiis/vllm-neuron-rhel9generic	*	Not reported
Red Hat/rhaiis/vllm-spyre-rhel9generic	*	Not reported
Red Hat/rhaiis/vllm-tpu-rhel9generic	*	Not reported
vLLM/vLLMgeneric	>=0.10.2 <0.13.0	0.13.0
vllmpip	>=0.10.2,<0.13.0	0.13.0
vllmpypi	>=0.10.2 <0.13.0	0.13.0

Recommended response

Check inventory. Check lockfiles and deployed manifests for Red Hat/rhaiis/vllm-neuron-rhel9, Red Hat/rhaiis/vllm-spyre-rhel9, Red Hat/rhaiis/vllm-tpu-rhel9.

Review the reported fix. Update vLLM/vLLM to 0.13.0; vllm to 0.13.0; vllm to 0.13.0 if you use the affected versions. Test the change in a non-production environment first.

Sources and claim methodology

Upstream source description

vLLM versions >= 0.10.2 and < 0.13.0 are missing sparse tensor validation in multimodal embeddings processing. Because PyTorch disables sparse tensor invariant checks by default, an attacker can submit crafted embedding requests with malformed (negative or out-of-bounds) tensor indices, when the prompt-embeds feature is enabled, to trigger crashes or resource exhaustion (denial of service), with potential for out-of-bounds/write-what-where memory corruption. This continues CVE-2025-62164, whose prior fix only disabled the feature by default rather than addressing the root cause.

Quoted source text, attributed separately from HOL analysis.

vLLM - Denial of Service via Unvalidated Multimodal EmbeddingsCVE-2026-56340

vLLM - Denial of Service via Unvalidated Multimodal EmbeddingsCVE-2026-56340

Key facts

Why this deserves its current priority

Affected scope and exposure questions

Recommended response

Evidence timeline and material changes

Sources and claim methodology