CVE-2026-114053

High

Description

vLLM from 0.7.3 before 0.31.0 forwards the request-supplied mm_processor_kwargs.code_revision to Transformers AutoProcessor.from_pretrained, letting clients select which revision of a model's dynamic processor code is imported when trust_remote_code is enabled. An attacker controlling a revision in the served model repository can send a multimodal request to the unauthenticated /tokenize endpoint to execute arbitrary Python in the vLLM API process.

SeverityHigh

CVSS8.1

Exploit probability—

ProductVllm

PublishedOct 21, 2026

First seenOct 21, 2026

Weakness typeCWE-94Code injection

How to fix

Update

vllm0.31.0 or later
vllm vllm0.31.0 or later

Patches

github.com/vllm-project/vllm/commit/570735520942b6428ae98f1eb442c6d88adae55e

Vendor advisories

github.com/vllm-project/vllm/security/advisories/GHSA-h3rc-6mm3-gc2m

Affected products

Product	Affected versions	Fixed in
vllmPyPI	0.7.3 ≤ v < 0.31.0	0.31.0
vllm vllm	0.7.3 ≤ v < 0.31.0	0.31.0

Check your version

Choose the product and enter the version you run to see whether it falls in the affected range.

In the Intelligence 360 plan

Attack Surface Management

See the assets tied to your domains, the technologies they run and known vulnerabilities in one list.