CVE-2026-115913
vLLM is an inference and serving engine for large language models. Prior to 0.32.0, the Phi-4 multimodal audio encoder in phi4mm_audio.py and phi4mm_utils.py uses eager attention by default, allowing a remote client to submit one legal long audio clip that causes MultiHeadedAttention.forward_attention to materialize and retain full attention matrices in self.attn. The scaled dot-product attention branch does not materialize these matrices. The resulting quadratic GPU-memory use can cause an out-of-memory condition and terminate EngineCore for all users. This issue is fixed in version 0.32.0.
Am I affected?
Sign in and add the software you use to your watchlist: we check it against every new vulnerability.
Show me how to fix itTimeline
- Published by NVD
- Last updated by NVD
Affected products
| Product | Versions |
|---|---|
| Vllm Vllm | from 0.31.0 to before 0.32.0 |
Sources: NVD and CISA known exploited vulnerabilities catalogue.