Inefficient Algorithmic Complexity Affecting vllm package, versions [0.10.1,0.31.0)


Severity

Recommended
0.0
high
0
10

CVSS assessment by Snyk's Security Team. Learn more

Threat Intelligence

EPSS
0.46% (41st percentile)

Do your applications use this vulnerable package?

In a few clicks we can analyze your entire application and see what components are vulnerable in your application, and suggest you quick fixes.

Test your applications
  • Snyk IDSNYK-PYTHON-VLLM-22549640
  • published14 Nov 2026
  • disclosed13 Nov 2026
  • creditJohn Q. Public

Introduced: 13 Nov 2026

NewCVE-2026-126397  (opens in a new tab)
CWE-407  (opens in a new tab)
CWE-770  (opens in a new tab)

How to fix?

Upgrade vllm to version 0.31.0 or higher.

Overview

vllm is an A high-throughput and memory-efficient inference and serving engine for LLMs

Affected versions of this package are vulnerable to Inefficient Algorithmic Complexity via Cohere2VisionProcessingInfo.get_num_patches, which forwards a request-controlled mm_processor_kwargs.max_patches value to the Hugging Face processor without applying a server-side upper bound. The processor's aspect-ratio enumeration in get_optimal_tiled_canvas performs two nested loops each bounded by the attacker-supplied value, resulting in quadratic CPU work during multimodal preprocessing. A remote client can submit small image chat-completion requests with large, varying max_patches values to bypass the LRU cache, exhaust preprocessing workers, and deny service to other clients.

Note: This is only exploitable when vLLM is serving a Cohere2 Vision model through the OpenAI-compatible frontend and the caller can submit image chat requests.

CVSS Base Scores

version 4.0
version 3.1