---
id: "CVE-2026-126394"
page: "https://infrastack.ai/live/CVE-2026-126394"
kind: "vulnerability"
aliases: ["CVE-2026-126394"]
title: "vLLM: unbounded frame count lets a video request crash the API server"
severity: "medium"
cvss: 6.9
epss: null
exploited: false
status: "open"
category: "ai-tooling"
products: []
ecosystems: []
tags: ["ai-tooling", "python", "self-hosted", "vllm"]
not_affected_if: ["no-public-api", "version-outside-range"]
published_at: "2026-11-13T10:17:43+00:00"
first_seen_at: "2026-11-13T10:47:18+00:00"
updated_at: "2026-11-13T10:54:47+00:00"
sources: [{"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-126394", "first_seen_at": "2026-11-13T10:47:18+00:00"}]
exploit_sources: []
---

# vLLM: unbounded frame count lets a video request crash the API server

Link for people: https://infrastack.ai/live/CVE-2026-126394 (this .md is for agents)

vLLM 0.30.0 before 0.31.0 has a video backend that takes the number of frames from the request without an upper limit. A crafted request with a huge frame count makes the server allocate large memory and use CPU until the API server is killed for running out of memory.

_Summary, check and fix written by a model from the sources; labelled as such._

## How to check

- Check the installed vLLM version with `pip show vllm` and see whether it is in `0.30.0` up to but not including `0.31.0`.
- Look for the openpangu video backend being selected (video_backend set to openpangu) in your serving configuration.

## Fix

- vLLM: move to `0.31.0` or later.

## Timeline

- 2026-11-13T10:47:18+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-126394

## Documents

- [advisory] Maintainer's advisory: github.com https://github.com/vllm-project/vllm/security/advisories/GHSA-894j-wg8m-63h8
- [advisory] Third-party advisory: www.vulncheck.com https://www.vulncheck.com/advisories/vllm-0.30.0-before-0.31.0-resource-exhaustion-via-openpangu-video-backend-num-frames

## From the sources

vLLM 0.30.0 before 0.31.0 contains a resource exhaustion vulnerability in OpenCVDynamicOpenPanguVideoBackend, which uses request-controlled num_frames without an effective ceiling. Unauthenticated attackers can send /v1/chat/completions or /tokenize requests selecting the openpangu video_backend with a huge num_frames, forcing np.linspace allocations and CPU work that OOM-kill the API server.

---

_Free to read, with fair use of about 60 pages a minute. For higher limits or API access: sales@infrastack.ai_