CVE-2026-115916

mediumpriority · track · 0EPSS 0.003 · 23 %ileDEUS
grimoire priorityP3 · Track
Exploitable?
—
no known exploitation
Likely / severe?
0.30% (23rd) EPSS
CVSS 6.5 · medium
Is it mine?
2 products
Check your exposure
Who says so?
2 sources
cert_bund · nvd

EPSS · scored 2026-10-28

vLLM is an inference and serving engine for large language models. Prior to 0.32.0, vllm/tokenizers/hf.py computes max_token_id with an off-by-one upper bound, allowing a pre-tokenized prompt ID equal to the vocabulary size to pass validation. For a tensor-parallel size of one and an exact multiple-of-64 vocabulary, VocabParallelEmbedding.forward does not mask that value before F.embedding uses it, causing a CUDA device-side assertion and terminating the engine. This issue is fixed in version 0.32.0.

01

cvss

Technical severity, 0–10. One row per scoring source.

6.5
010

base score 0–10 · higher = more severe · hover a vector to decode

v3.16.5mediumCVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:Hsecondary · security-advisories@github.comexploitability 2.8 · impact 3.6
v3.16mediumCVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H/E:Usecondary · CVSS-BT (t0sche)
02

weaknesses

The CWE weakness classes this flaw belongs to.

03

affected products (2)

Products and version ranges known to be affected.

04

advisories (1)

Vendor and ecosystem advisories describing this.

  • WID-SEC-W-2026-4051vllm: Mehrere SchwachstellenDeutsch2026-10-19csaf · cert_bundDE

    Ein Angreifer kann mehrere Schwachstellen in vllm ausnutzen, um beliebigen Programmcode auszuführen, um einen Denial of Service Angriff durchzuführen, um Informationen offenzulegen und um Sicherheitsvorkehrungen zu umgehen.

05

sources

Which sources assert this record, and when.

published 2026-10-26modified 2026-10-28
07

references (4) nvd

External citations — patches, advisories, analysis.

scry

Attackers weaponise vulnerabilities like this against internet-exposed services. See what is exposed on your perimeter with a free passive scan.

Free scan