Skip to content
Threat Brief

← All advisories

CVE-2026-130686

High 8.7No known exploitationFix available

vLLM 0.6.0 before 0.33.0 Resource Exhaustion via Multi-Image Requests

Record last updated

CVSS score

8.7

High on a 0 to 10 scale

EPSS, chance of exploitation in 30 days

0.4%

Higher than 36% of scored CVEs. A prediction, not evidence.

CISA remediation deadline

None

Not in the KEV catalog

Description

vLLM from 0.6.0 before 0.33.0 lacks a request-wide media budget, allowing remote inference clients to exhaust frontend resources by submitting many remote image URLs in one request. Attackers can send a compact request to /v1/messages or /v1/responses that forces downloading and decoding of every image before prompt-length admission, exhausting memory, CPU, and connections.

Weakness: CWE-400

Patching and mitigation

Patch status
Fix available (source: CVE record (VulnCheck))
Fix documented (source date)
Mitigation
None documented
Vendor guidance
github.com (opens in a new tab)

Affected products

VendorProductAffected versionsFixed inPer
vllm-projectvllmfrom 0.6.0 before 0.33.00.33.0CVE Program
vllm-projectvllm (unaffected)0.33.0Not documentedCVE Program

Timeline

  1. Disclosure

    CVE record published by VulnCheck

    Source: CVE Program (opens in a new tab)

  2. Patch

    Fix documented by CVE record (VulnCheck)

    Source: CVE record (VulnCheck) (opens in a new tab)

Vendor guidance and references