5.3

CVE-2026-105758: vllm-project vllm

CVSS: 5.3v3.1MediumAnalyzed
AliasesGHSA-x6mc-67gf-chw4CGA-3gvr-6hg2-7j3pCGA-5rw2-62r4-fhm4
Published 05-10-2026Last modified 08-10-2026

Vulnerability summary

CVE-2026-105758 is a medium (CVSS 5.3) vulnerability in vllm-project vllm. EPSS estimates a 0.30% chance of exploitation in the next 30 days, higher than 21.8% of scored vulnerabilities.

Overall risk
21/100
0100
Low · default modelHow is this calculated?
CVSS score
5.3/10
010
Medium · v3.1 (CVE.org)
EPSS score
0.30%
0%100%
Low · Change unavailable
KEVNot listedIn no KEV catalog
RansomwareNone recordedNo known use recorded
Public exploits0No public exploits recorded

Timeline

  1. 05-10-2026
    PublishedVulnerability first disclosed.
  2. 08-10-2026
    Last modifiedVulnerability information updated.

Description

vLLM is an inference and serving engine for large language models. From 0.24.0 until 0.30.0, the Qwen2VLVideoBackend and Qwen3VLVideoBackend classes accept request-level values for the media_io_kwargs.video.max_frames and media_io_kwargs.video.fps fields without enforcing server-side ceilings. An unauthenticated caller can submit these values to the /tokenize endpoint, causing the sampler to decode every frame selected from attacker-controlled video input, consume disproportionate frontend memory, and potentially terminate the API process before scheduling or admission control. The Rust frontend is not affected because it rejects the media_io_kwargs field. This issue is fixed in version 0.30.0.

CVSS metrics

4 scores · 4 sources · 1 version

EPSS trend

Probability

0.30%

Chance of exploitation in the next 30 days

Percentile

22%

Versus all other CVEs

30-day change

Change unavailable

Change unavailable

06-10-202621-10-2026
EPSS score
0.30%

Known exploited

Compare KEV catalogs

Not in any of the 3 tracked catalogs.

Public exploits

0 found

No public exploits indexed.

Weaknesses

CWE · root cause

Affected systems

Chainguard4 products
  • kserve-huggingfaceserver-cuda-13.3All versions (No fix recorded)
  • tritonserver-backend-vllm-cuda-13.0All versions (No fix recorded) | < 25.11-r17
  • vllm-cuda-13.2All versions (No fix recorded) | < 0.32.0-r0
  • vllm-openai-cuda-13.0All versions (No fix recorded) | < 0.31.0-r0
PyPI1 product
  • vllm≥ 0.24.0, < 0.30.0
Vllm1 product
  • vllm≥ 0.24.0, < 0.30.0
Vllm-project1 product
  • vllm≥ 0.24.0, < 0.30.0

4 vendors · 7 products

References

9 links