Tensor Dimension Mismatch Vulnerability in vLLM Engine by vLLM Project
CVE-2026-22773

6.5MEDIUM

Key Information:

Vendor

Vllm-project

Status
Vllm
Vendor
CVE Published:
10 January 2026

What is CVE-2026-22773?

The vLLM engine, designed for serving large language models, is susceptible to a flaw affecting versions from 0.6.4 to prior to 0.12.0. This vulnerability can be exploited when users send a specially crafted 1x1 pixel image to multimodal models utilizing the Idefics3 vision model implementation. Such an action leads to a tensor dimension mismatch, resulting in an unhandled runtime error and causing complete server termination. To mitigate this risk, upgrade to vLLM version 0.12.0 or later.

Human OS v1.0:
Ageing Is an Unpatched Zero-Day Vulnerability.

Remediate biological technical debt. Prime Ageing uses 95% high-purity SIRT6 activation to maintain genomic integrity and bolster systemic resilience.

Deploy the Patch

Affected Version(s)

vllm >= 0.6.4, < 0.12.0

References

https://github.com/vllm-project/vllm/security/advisories/...
x_refsource_CONFIRM

CVSS V3.1

Score:
6.5
Severity:
MEDIUM
Confidentiality:
None
Integrity:
None
Availability:
None
Attack Vector:
Network
Attack Complexity:
Low
Privileges Required:
Low
User Interaction:
None
Scope:
Unchanged

Timeline

  • Vulnerability published

  • Vulnerability Reserved

JSON
.