Resource exhaustion in Vllm-Project Vllm
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.
Published · last modified .
CVSS v3 metric
CVSS v3 base score 6.5 (Medium). Vector: CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H.
EPSS exploit prediction
No EPSS score from FIRST.org yet. New CVEs usually get one within a few days; no score means no signal yet, not low risk. How EPSS works.
Affected products
- Vllm-Project Vllm — versions < 0.32.0
Weakness classification (CWE)
References
Frequently asked questions
- What is CVE-2026-115913?
- CVE-2026-115913 is a medium-severity vulnerability in Vllm-Project Vllm, classified under Allocation of Resources Without Limits or Throttling. CVSS score: 6.5/10. Published 2026-10-26.
- How severe is CVE-2026-115913?
- Medium severity. CVSS v3 base score is 6.5 out of 10.