CVE-2026-44223

medium
Published 2026-05-12 · Modified 2026-05-20
CVSS v3
6.5
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H
CVSS v2
VIR risk
6.5

Description

vLLM is an inference and serving engine for large language models (LLMs). From to before 0.20.0, the extract_hidden_states speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The crash is triggered when any request in the batch uses sampling penalty parameters (repetition_penalty, frequency_penalty, or presence_penalty). A single request with a penalty parameter (e.g., "repetition_penalty": 1.1) is sufficient to crash the server. This vulnerability is fixed in 0.20.0.

Predictions

Exploit likelihood
75%
Patch ETA

Heuristic predictions, AS-IS, for prioritization only.

Mitigations

vendor Authored 2026-05-27

Vendor advisory: security-advisories@github.com — https://github.com/vllm-project/vllm/security/advisories/GHSA-83vm-p52w-f9pw

vendor Authored 2026-05-27

Vendor advisory: security-advisories@github.com — https://github.com/vllm-project/vllm/pull/38610

Package impact

EcosystemPackageVulnerableFixed
python PyPIvllm>=0.18.0,<0.20.00.20.0
PIPvllm>= 0.18.0, < 0.20.00.20.0

Application impact

VendorProductVersionsFixed
vllmvllm{"startIncluding":"0.18.0","endExcluding":"0.20.0"}0.20.0

References

CWEs

CWE-131 CWE-704

Verify integrity in audit chain (admin only). AS-IS.