Skip to main content
CVE-2026-116178VulnerabilityPublished 10/25/2026

CVE-2026-116178: vllm · Improper Input Validation

Affects vllm-project vllmSource CVE Database V5

0

Not known to be exploited · CVSS 8.2 High.

Data as of 10/25/2026, 16:50:55 UTC

CVSS 4.0 base score
8.2High
How severe the flaw is on the 0–10 CVSS scale.
CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N

Exploitation evidence

  1. Disclosed: evidence found.

    CVE published 25 Oct 2026

  2. Detectable: no evidence on record.

    No public nuclei template indexed

  3. Weaponised: no evidence on record.

    No public exploit code indexed in Metasploit or Exploit-DB

  4. Exploited in the wild: no evidence on record.

    Not in CISA KEV and no source reports exploitation

  5. Observed by OffSeq Mirage: no evidence on record.

    No attack traffic recorded by OffSeq Mirage sensors

A missing step is not evidence of safety. Sources: cve.org (opens in a new tab), the Metasploit, Exploit-DB and nuclei-templates index as of 2026-10-25.

What to do

Affected software

ProductAffectedFixed in
vllm-projectvllmAffected
  • >=0.20.0 <0.30.0
Fixed in
  • 0.30.0
GitHubvllm-project/vllmpkg:github/vllm-project/vllmOther threatsAffected–Fixed in–

Check your own hosts for these packages with threat-finder (opens in a new tab), our free open-source scanner.

AI-suggested

Until a fix is available, avoid processing untrusted or malformed multimodal metadata inputs to the /inference/v1/generate endpoint.

Description

vLLM versions from 0.20.0 up to but not including 0.30.0 contain a denial of service vulnerability. This flaw allows remote attackers to crash the inference engine by sending malformed multimodal metadata to the /inference/v1/generate endpoint. The issue arises from improper input validation, specifically when replaying valid multimodal items with repeated or overlapping placeholders or mismatched data lists, leading to a negative text span in Qwen2.5-VL position construction that terminates the engine.

Impact

Successful exploitation results in a denial of service condition by crashing the vLLM inference engine remotely. This disrupts availability of the service but does not indicate code execution or data compromise based on the provided data.

References

Community reviews

Loading reviews…