# CVE-2026-105757 — vllm-project vllm

> As of 2026-10-06, CVE-2026-105757 is a MEDIUM-severity vulnerability in vllm-project vllm, CVSS v3.1 6.5, EPSS 0.3% (22.0th percentile). No Threadlinqs-tracked threat campaign has been attributed to CVE-2026-105757 as of 2026-10-06; the identifier is re-checked against the Threadlinqs threat corpus on every daily ingest.

**Last updated:** 2026-10-06

## What is CVE-2026-105757?

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, structured-output request failures can escape request-scoped validation and reach the EngineCore fatal-error path. A per-request backend mismatch can re-raise a grammar compilation exception, padding produced by the ngram_gpu speculative-decoding mode can pass a negative token to guidance validation, and the Rust frontend can admit empty structured-output values that the Python frontend rejects, allowing ordinary constrained-generation requests to terminate the shared engine. This issue is fixed in version 0.30.0.

The record classifies CVE-2026-105757 under weakness classes [CWE-20](https://intel.threadlinqs.com/cwe/CWE-20), [CWE-248](https://cwe.mitre.org/data/definitions/248.html), [CWE-755](https://cwe.mitre.org/data/definitions/755.html). Its CVSS v3 base vector states that the flaw is reachable remotely over the network, needs low-privilege credentials, needs no user interaction, and has high impact on availability. 1 affected-product entry is recorded, across 1 vendor, listed below. The identifier was first published 45 days ago.

## Severity and exploitation probability

- **CVSS v3.1 base score:** 6.5 — MEDIUM (`CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H`)
- **EPSS (FIRST):** 0.3% probability of exploitation in the next 30 days, higher than 22.0% of all scored CVEs
- **CISA KEV:** Not listed in the CISA Known Exploited Vulnerabilities catalog
- **Threadlinqs priority:** 4.1/10 — a Threadlinqs composite of the CVSS base score, the EPSS percentile and public exploit availability
- **Published:** 2026-10-05

## Is CVE-2026-105757 being exploited?

It currently carries a trending score of 35 in the Threadlinqs vulnerability feed.

## Affected products and versions

- [vllm-project](https://intel.threadlinqs.com/vendors/vllm-project): vllm

## How to fix CVE-2026-105757

No vendor patch reference has been recorded for CVE-2026-105757 in the tracked sources. Follow the references below for a fix, and treat the products listed above as exposed until the vendor states otherwise.

## Threat activity tracking CVE-2026-105757

No threat campaign in the Threadlinqs corpus currently references CVE-2026-105757, in its CVE list or as an indicator. The linkage is recomputed from the live corpus on every daily ingest, so this section fills in as soon as a tracked campaign cites the identifier.

## Sources

Enriched from CVE.org, FIRST EPSS, NVD. This product uses the NVD API but is not endorsed or certified by the NVD.

**Vendor advisory and patch**

- [github.com](https://github.com/vllm-project/vllm/security/advisories/GHSA-85xf-c7hm-whqw)

**Other references**

- [github.com](https://github.com/vllm-project/vllm/pull/51450)
- [github.com (c55e15a44ec4127832d4a86928a356fdd9e68dbd)](https://github.com/vllm-project/vllm/commit/c55e15a44ec4127832d4a86928a356fdd9e68dbd)
- [github.com (v0.30)](https://github.com/vllm-project/vllm/releases/tag/v0.30.0)

Canonical: https://intel.threadlinqs.com/cve/CVE-2026-105757
No Threadlinqs detection currently covers CVE-2026-105757; it is re-checked on every ingest.
Indicator values unlock with TL Creds (https://threadlinqs.com/tl-creds); MCP access for AI agents needs Purple: https://intel.threadlinqs.com/mcp
