---
id: "CVE-2026-116181"
kind: "vulnerability"
aliases: ["CVE-2026-116181"]
title: "vLLM: out-of-vocabulary stop_token_ids crash the engine remotely"
severity: "high"
cvss: 7.1
epss: null
exploited: false
status: "open"
category: "ai-tooling"
products: []
ecosystems: []
tags: ["ai-tooling", "llm-serving", "python", "self-hosted", "vllm"]
not_affected_if: ["default-config"]
published_at: "2026-10-25T14:16:57+00:00"
first_seen_at: "2026-10-25T14:41:23+00:00"
updated_at: "2026-10-25T14:42:35+00:00"
sources: [{"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-116181", "first_seen_at": "2026-10-25T14:41:23+00:00"}]
---

# vLLM: out-of-vocabulary stop_token_ids crash the engine remotely

vLLM versions from 0.20.1 before 0.29.0 fail to validate stop_token_ids in requests, so an out-of-vocabulary ID can crash the inference engine. The same malformed input can also write outside the V2 runner's logits buffer.

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

## How to check

- Check the installed version with `pip show vllm` or `python -c "import vllm; print(vllm.__version__)"`
- Confirm whether the OpenAI-compatible server is reachable from untrusted networks, since the sources say remote attackers can send chat, completion, generate or batch requests

## Fix

- Upgrade vLLM to `0.29.0` or later
- Until upgraded, restrict network access to the vLLM API server

## Timeline

- 2026-10-25T14:41:23+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-116181

## From the sources

vLLM from 0.20.1 before 0.29.0 contains an improper input validation vulnerability that allows remote attackers to crash the engine by supplying out-of-vocabulary stop_token_ids. Attackers can send OpenAI chat, completion, generate, or batch requests with invalid IDs to terminate the engine or write outside the V2 runner's logits buffer.

---

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