Live › vulnerability › CVE-2026-115914
Go to dashboardvLLM versions from 0.11.1 up to 0.32.0 let a remote client send a specific legal image to the DeepSeek-OCR image path, which makes the serving engine hit a fatal assertion and shut down. An attacker can take the engine offline for every user of the server with one request.
Summary written from the sources below. Check the originals before acting.
First seen by InfraStack 2026-10-26 18:31 UTC. All times on this page are UTC.
Teams running vLLM between 0.11.1 and 0.32.0 that serve DeepSeek-OCR, DeepSeek-OCR-2 or Unlimited-OCR models and accept image input from clients that can reach the OpenAI-compatible API. Deployments that serve only text models or keep the API behind a private network are less exposed. Teams on 0.32.0 or later can skip it.
No product ranges from the sources yet.
python -c "import vllm; print(vllm.__version__)" to read the installed vLLM version0.32.0 or laterEach row is the first time that source carried this signal. The analyst’s rows are its own findings.
vLLM is an inference and serving engine for large language models. From 0.11.1 until 0.32.0, the DeepSeek-OCR, DeepSeek-OCR-2, and Unlimited-OCR image path allows a remote client that can reach the OpenAI-compatible API to submit a legal image whose normalized pixels sum to zero when that image is alone in its per-step encoder batch. _parse_and_validate_image_input returns no embedding, embed_multimodal propagates that value, and sanity_check_mm_encoder_outputs triggers a fatal assertion that shuts down the engine for all users. This issue is fixed in version 0.32.0.
This page as .md · .json · in your agent: get_signal("CVE-2026-115914")
NVD data is provided by NIST and is public domain. Posts quoted belong to their authors.