---
id: "CVE-2026-116180"
kind: "vulnerability"
aliases: ["CVE-2026-116180"]
title: "vLLM: crafted prompt_logprobs request crashes the engine and drops all requests"
severity: "high"
cvss: 8.3
epss: null
exploited: false
status: "open"
category: "ai-tooling"
products: []
ecosystems: []
tags: ["ai-tooling", "llm-serving", "python", "self-hosted", "vllm"]
not_affected_if: ["no-kv-connector", "single-tenant", "version-0.31.0-or-later"]
published_at: "2026-10-25T14:16:57+00:00"
first_seen_at: "2026-10-25T14:41:23+00:00"
updated_at: "2026-10-25T14:42:22+00:00"
sources: [{"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-116180", "first_seen_at": "2026-10-25T14:41:23+00:00"}]
---

# vLLM: crafted prompt_logprobs request crashes the engine and drops all requests

A flaw in vLLM's V1 engine lets a client that shares the engine send a request with prompt_logprobs (or echo with logprobs) against a KV-connector-cached prefix. The resulting tokenizer OverflowError crashes EngineCore and drops every co-tenant request.

_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__)"`; versions from 0.8.5 up to but not including 0.31.0 are listed as affected.
- Check whether the deployment uses a KV connector (the V1 engine with a KV transfer config) and accepts requests with prompt_logprobs or echo enabled from untrusted clients.

## Fix

- Upgrade vLLM to `0.31.0` or later.
- Until upgraded, reject or strip prompt_logprobs and echo-with-logprobs requests at the gateway in front of vLLM.

## Timeline

- 2026-10-25T14:41:23+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-116180

## From the sources

vLLM from 0.8.5 before 0.31.0 has a denial of service flaw. LogprobsTensors.empty_cpu() in vllm/v1/outputs.py allocates an uninitialized int32 tensor, and the V1 model runner leaves prompt positions reported as computed by a KV connector unwritten. A client sharing the engine can send prompt_logprobs (or echo with logprobs) against a connector-cached prefix, making tokenizer decode raise OverflowError, which crashes EngineCore and drops all co-tenant requests.

---

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