# CVE-2026-116178: vLLM denial of service in multimodal inference endpoint

Source: https://junglewise.ai/threats/cve-2026-116178-vllm-from-0-20-0-before-0-30-0-contains-a-denial-of-service  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-10-25. Updated: 2026-10-25  
License: CC BY 4.0 with a link back

CVE-2026-116178 · Severity: medium · CVSS 5.9 · Published 2026-10-25

vLLM is a large language model serving engine used to run AI models in production. Versions 0.20.0 through 0.29.x contain a flaw in the multimodal inference endpoint that allows remote attackers to crash the engine by sending malformed image/media metadata, disrupting service for all users until the engine restarts.

Technologies: [vllm (PyPI)](https://junglewise.ai/threats/technologies/pypi-vllm), [vLLM Project vllm](https://junglewise.ai/threats/technologies/vllm-project-vllm). Vendors: [PyPI](https://junglewise.ai/threats/vendors/pypi), [vLLM Project](https://junglewise.ai/threats/vendors/vllm-project).

## Quick answers

**Is CVE-2026-116178 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-116178 being exploited in the wild.

**How severe is CVE-2026-116178?** Medium severity, with a CVSS score of 5.9, published 25 October 2026.

**Which products and versions are affected?** vllm (PyPI) ECOSYSTEM:introduced=0.20.0,fixed=0.30.0; vLLM Project vllm 0.20.0 to 0.29.x.

**Is there a fix for CVE-2026-116178?** Yes. vllm (PyPI): upgrade to 0.30.0 or later. Until then: Restrict network access to the /inference/v1/generate endpoint using firewall rules or API gateway authentication until upgrading to 0.30.0 or later. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.

**How is CVE-2026-116178 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-116178?** Check whether you run vllm (PyPI) or vLLM Project vllm at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## How to fix

- vllm (PyPI): upgrade to 0.30.0 or later

  ```sh
  pip install "vllm>=0.30.0"
  ```


Restrict network access to the /inference/v1/generate endpoint using firewall rules or API gateway authentication until upgrading to 0.30.0 or later.

## Technical details

The disaggregated multimodal /inference/v1/generate endpoint fails to validate client-supplied kwargs_data, mm_hashes, and mm_placeholders fields before passing them to model code. An attacker can send repeated or overlapping placeholder ranges or mismatched list lengths that violate assumptions in model position construction (e.g., Qwen2.5-VL), triggering a fatal crash. This is a network-accessible input validation flaw (CWE-20) requiring no authentication or user interaction.

## Affected products

- vLLM project vLLM 0.20.0 to 0.29.x

## Identifiers

- CVE-2026-116178
- GHSA-34qw-q2h2-jp3m
- PYSEC-2026-4353

## Timeline

- 2026-10-09: disclosed
- 2026-10-09: patched: fix released in v0.30.0
- 2026-10-25: advisory

## References

- [https://github.com/vllm-project/vllm/security/advisories/GHSA-34qw-q2h2-jp3m](https://github.com/vllm-project/vllm/security/advisories/GHSA-34qw-q2h2-jp3m)
- [https://www.vulncheck.com/advisories/vllm-0.20.0-before-0.30.0-dos-via-inference-v1-generate-mm-placeholders](https://www.vulncheck.com/advisories/vllm-0.20.0-before-0.30.0-dos-via-inference-v1-generate-mm-placeholders)

## Related threats

- [vLLM chat template resource exhaustion DoS](https://junglewise.ai/threats/cve-2026-115918-vllm-is-an-inference-and-serving-engine-for-large-language)
- [vLLM: Permissive default CORS origins allow cross-origin requests to the vLLM HTTP server](https://junglewise.ai/threats/cve-2026-115917-vllm-is-an-inference-and-serving-engine-for-large-language)
- [vLLM off-by-one error in tokenizer max_token_id validation](https://junglewise.ai/threats/cve-2026-115916-vllm-is-an-inference-and-serving-engine-for-large-language)
- [vLLM image format validation bypass in media processing](https://junglewise.ai/threats/cve-2026-115915-vllm-is-an-inference-and-serving-engine-for-large-language)
- [vLLM denial of service in DeepSeek-OCR image processing](https://junglewise.ai/threats/cve-2026-115914-vllm-is-an-inference-and-serving-engine-for-large-language)