---
id: "CVE-2026-116177"
kind: "vulnerability"
aliases: ["CVE-2026-116177"]
title: "vLLM: crafted image triggers memory exhaustion, denying service to other users"
severity: "high"
cvss: 8.7
epss: null
exploited: false
status: "open"
category: "ai-tooling"
products: []
ecosystems: []
tags: ["ai-inference", "ai-tooling", "python", "self-hosted", "vllm"]
not_affected_if: ["no-nemotron-vl-models", "text-only-serving"]
published_at: "2026-10-25T14:16:57+00:00"
first_seen_at: "2026-10-25T14:41:23+00:00"
updated_at: "2026-10-25T14:42:17+00:00"
sources: [{"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-116177", "first_seen_at": "2026-10-25T14:41:23+00:00"}]
---

# vLLM: crafted image triggers memory exhaustion, denying service to other users

vLLM before 0.25.0 lets unauthenticated users submit a decompression-bomb image to the chat completions endpoint, because the Nemotron-VL processors globally disable Pillow's pixel limit. The oversized image exhausts memory, OOM-kills the worker and takes the service down for co-tenants.

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

## How to check

- Run `pip show vllm` (or `python -c "import vllm; print(vllm.__version__)"`) and check whether the version is between `0.10.2` and `0.25.0`
- Check whether the deployed model is a Nemotron-VL model, since the flaw is in those processors
- unknown

## Fix

- Upgrade vLLM to `0.25.0` or later
- Until upgraded, limit request body size and memory per worker at the gateway or container level

## Timeline

- 2026-10-25T14:41:23+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-116177

## From the sources

vLLM from 0.10.2 before 0.25.0 contains an uncontrolled resource consumption vulnerability that allows unauthenticated attackers to exhaust memory because the Nemotron-VL processors globally set Image.MAX_IMAGE_PIXELS to None. Attackers can submit a pathological decompression-bomb image to /v1/chat/completions to OOM-kill the worker and deny service to co-tenants.

---

_Free to read at a human pace. An API with keys and an MCP server are in preparation; early access: hello@infrastack.ai_