Package
py3.12-vllm-cuda-12.4
Component
vllm
Latest update
6.5
CVSS V3
Build, ship, and run secure software with minimal, hardened container images — rebuilt from source daily and guarded under our industry-leading remediation SLA.
Start for freeStatus
Justification
Impact
vllm 0.18.1 does not contain the PyNvVideoCodec GPU video decoding backend this CVE affects. The v0.18.1 tree has no pynvvideocodec references in vllm/multimodal/video.py, vllm/multimodal/media/video.py or vllm/v1/worker/gpu_worker.py; the only registered video loaders are opencv, opencv_dynamic, molmo2, nemotron_vl and openpangu, all CPU decoders. The PyNvVideoCodec backend first appears in vLLM 0.25.0, so a request selecting it cannot reach a GPU decoder here. The vulnerable code is absent from this package.
Status