CVE-2026-105757: vLLM EngineCore Denial of Service via Structured-Output Request Errors
Summary
CVE-2026-105757: Affected - Ecosystem / package: pip / vllm - Affected versions: vLLM ≤ 0.25.1 (confirmed on 0.25.1, commit 752a3a504485). 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 serving. It allows users to make API calls for structured outputs, enabling advanced functionalities in machine learning applications. However, certain versions of vLLM are susceptible to vulnerabilities that can lead to denial of service, affecting the availability of the service for all users sharing the same EngineCore.
Vulnerability Detail
The vulnerability arises from three structured-output request paths that allow an ordinary request to trigger an uncaught, engine-fatal exception. This exception escapes the request boundary and affects the shared EngineCore, leading to a denial of service for all concurrent tenants. The root cause is the lack of a per-request exception boundary around structured-output grammar and token handling, which allows a malformed request to propagate as a fatal error. This vulnerability can be exploited by any client capable of sending structured-output requests, resulting in service unavailability without requiring privileged access.
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, you can implement input validation on structured-output requests to ensure that they conform to expected formats before they reach the EngineCore. However, this is not a substitute for applying the patch.
Conclusion
CVE-2026-105757 is a significant vulnerability in the vLLM package that can lead to denial of service for applications relying on shared EngineCore instances. It is crucial for users to update to the patched version to mitigate this risk and ensure the continued availability of their services.
See every vulnerability recorded for vllm →
FAQ
CVE-2026-105757 is a vulnerability in the vLLM package that allows structured-output request errors to escape their request boundaries, leading to a denial of service in the shared EngineCore.
You can check if your application is affected by using the Vulert playground at Vulert Scanner, which monitors and alerts for open-source vulnerabilities in real-time.
Versions of vLLM up to and including 0.25.1 are affected by this vulnerability. It is recommended to upgrade to version 0.30.0 or later.
The primary impact is denial of service, which means that any client can terminate the shared EngineCore, affecting all users relying on that engine instance.
No, the vulnerability primarily affects availability and does not lead to data loss, code execution, or memory corruption.
References
Related vulnerabilities
Recent vulnerabilities recorded for vllm.
