CVE-2026-34760
vLLM is an inference and serving engine for large language models (LLMs).
Does this matter?
High impact if exploited, but EPSS currently rates exploitation as unlikely (0.27%). Schedule it in the normal patch cycle and watch for a rise in EPSS or a public exploit.
Description
vLLM is an inference and serving engine for large language models (LLMs). From version 0.5.5 to before version 0.18.0, Librosa defaults to using numpy.mean for mono downmixing (to_mono), while the international standard ITU-R BS.775-4 specifies a weighted downmixing algorithm. This discrepancy results in inconsistency between audio heard by humans (e.g., through headphones/regular speakers) and audio processed by AI models (Which infra via Librosa, such as vllm, transformer). This issue has been patched in version 0.18.0.
- CVSS 3.1
- 7.1 HIGHCVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:L
- EPSS
- 0.27% probability · 19th percentile
- CISA KEV
- Not listed
- Weakness
- CWE-20
- Affected
- vllm/vllm
- Source
- security-advisories@github.com
References
- https://github.com/vllm-project/vllm/commit/c7f98b4d0a63b32ed939e2b6dfaa8a626e9b46c4Patch
- https://github.com/vllm-project/vllm/pull/37058Issue Tracking
- https://github.com/vllm-project/vllm/releases/tag/v0.18.0Release Notes
- https://github.com/vllm-project/vllm/security/advisories/GHSA-6c4r-fmh3-7rh8Vendor Advisory
Source: NVD record, EPSS from FIRST.org, KEV from CISA. Refreshed daily.