CVE-2026-94623: vLLM Denial of Service Vulnerability in NIXL Connector
Summary
CVE-2026-94623: vLLM through 0.29.0 contains a denial of service vulnerability in the NIXL connector's prefix caching implementation that fails to properly validate block counts across multi-prompt completion requests in prefill/decode... Affects vllm versions < 0.30.0. Fixed in 0.30.0. How to fix ↓
Check if this vulnerability affects your application
Upload a manifest or SBOM and we'll show whether you're affected, plus every other risky package in the same scan.
No signup required · Results in about 60 seconds
Background
vLLM is an open-source library designed for efficient and scalable language model inference. It provides various connectors, including NIXL, to facilitate multi-prompt completion requests. The library is widely used in applications requiring natural language processing capabilities. However, a vulnerability has been identified in the NIXL connector's prefix caching implementation, which can lead to service disruptions.
Vulnerability Detail
The vulnerability arises from improper validation of block counts during multi-prompt completion requests in disaggregated deployments. Attackers can exploit this flaw by submitting completion requests with multiple prompts of varying lengths, triggering an assertion failure in the NixlBaseConnectorWorker._apply_prefix_caching method. This results in the decode worker terminating unexpectedly, rendering it unavailable until a restart occurs. The potential ramifications include denial of service, where legitimate users cannot access the service due to the worker's unavailability.
How To Fix
Latest published version: 0.32.0 as of 22 Oct 2026
pip install --upgrade "vllm==0.32.0"Upgrading to it may include breaking changes; the version below is the minimum fixed release for this vulnerability.
If you need the minimum change
0.30.0
This is the first release that contains the fix.
pip install --upgrade "vllm==0.30.0"Tip — Not sure which version your app runs? Vulert can match this fix to it. Scan free →
Workaround
As a temporary workaround, users can limit the number of prompts sent in a single completion request to avoid triggering the vulnerability. This can be done by modifying the request logic in your application to ensure that it adheres to a single prompt per request until the package is updated.
Conclusion
The vLLM library has a critical denial of service vulnerability in its NIXL connector, which can be exploited through multi-prompt completion requests. Users are strongly advised to update to version 0.30.0 or later to mitigate this risk. Temporary workarounds are available, but updating is the most effective solution.
See every vulnerability recorded for vllm →
FAQ
CVE-2026-94623 is a denial of service vulnerability found in the vLLM library's NIXL connector, which can cause service interruptions.
You can use the Vulert playground at Vulert Scanner to check if your application is affected by this vulnerability without needing to sign up.
Exploitation of this vulnerability can lead to denial of service, making the decode worker unavailable until it is restarted.
You can update the vLLM package by running the command pip install --upgrade vllm in your terminal.
Yes, you can limit the number of prompts in your completion requests to one to avoid triggering the vulnerability until you can update the package.
Related vulnerabilities
Recent vulnerabilities recorded for vllm.
