Advisory

PYSEC-2026-145

Also known as: CVE-2026-44223, GHSA-83vm-p52w-f9pw

Summary

The record has no summary.

Details

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.

Severity

Medium Basis: CVSS v3 score 6.5, vector CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H

Dates

Published
2026-05-12 20:16 UTC
Modified
2026-05-20 09:19 UTC

Affected packages and fixes

vllm

PyPI · All systems

Affected

  • introduced 0.18.0, fixed 0.20.0 (ecosystem)

Fixed versions

0.20.0

Fix

  • Fixed version 0.20.0
    pip install --upgrade vllm==0.20.0

Fix commands are the export's own generic commands. Ojuju never runs them. Check them against your setup first.

Check your own computer: guard packages check

View this record on OSV.dev

Source

Source: PyPA Advisory Database · CC BY 4.0

Sources & licences