# CVE-2026-105753: vLLM multimodal cache assertion failure and availability impact

Source: https://junglewise.ai/threats/cve-2026-105753-vllm-is-an-inference-and-serving-engine-for-large-language  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-10-05. Updated: 2026-10-05  
License: CC BY 4.0 with a link back

CVE-2026-105753 · Severity: medium · CVSS 6.5 · Published 2026-10-05

vLLM is an inference and serving engine for large language models that processes multimodal requests (text, images, video, etc.). A flaw in the default caching layer causes the service to crash with an assertion error when processing repeated media requests, resulting in denial of service for all users sharing that inference engine.

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

## Quick answers

**Is CVE-2026-105753 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-105753 being exploited in the wild.

**How severe is CVE-2026-105753?** Medium severity, with a CVSS score of 6.5, published 5 October 2026.

**Which products and versions are affected?** vllm (PyPI) < 0.28.0, ECOSYSTEM:introduced=0,fixed=0.28.0; Vllm prior to 0.28.0.

**Is there a fix for CVE-2026-105753?** Yes. vllm (PyPI): upgrade to 0.28.0 or later. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.

**How is CVE-2026-105753 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-105753?** Check whether you run vllm (PyPI) or 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 0.28.0 or later

  ```sh
  pip install "vllm>=0.28.0"
  ```


## Technical details

The multimodal LRU cache commits a media hash in the frontend sender cache before the engine layer admits the request; if the request is rejected, the receiver cache never receives the payload. A subsequent request reusing the same media hash causes the sender to skip payload transmission while the receiver hits an assertion "Expected a cached item," crashing the service. This is a cache coherency bug affecting shared inference deployments.

## Affected products

- vLLM vLLM prior to 0.28.0

## Identifiers

- CVE-2026-105753
- GHSA-ph3r-5jfg-f84f
- PYSEC-2026-4197

## Timeline

- 2026-10-05: disclosed
- 2026-08-12: patched: Fix merged in commit 396204230423b7cc6798300926b8fa30190d26a9

## References

- [https://github.com/vllm-project/vllm/commit/396204230423b7cc6798300926b8fa30190d26a9](https://github.com/vllm-project/vllm/commit/396204230423b7cc6798300926b8fa30190d26a9)
- [https://github.com/vllm-project/vllm/pull/46747](https://github.com/vllm-project/vllm/pull/46747)

## 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)