CVE-2026-115916

vLLM is an inference and serving engine for large language models. Prior to 0.32.0, vllm/tokenizers/hf.py computes max_token_id with an off-by-one upper bound, allowing a pre-tokenized prompt ID equal to the vocabulary size to pass validation. For a tensor-parallel size of one and an exact multiple-of-64 vocabulary, VocabParallelEmbedding.forward does not mask that value before F.embedding uses it, causing a CUDA device-side assertion and terminating the engine. This issue is fixed in version 0.32.0.

Score

CVSS 3.1 6.5 Medium

Status

Analyzed - security-advisories@github.com

Published

2026-10-26

Last modified

2026-10-28

Weaknesses

CWE-129, CWE-193

Affected products

• Vllm (Vllm)
• Vllm (Vllm-project)

References

• https://github.com/vllm-project/vllm/security/advisories/GHSA-q97c-p368-75c4
• https://github.com/vllm-project/vllm/pull/59491
• https://github.com/vllm-project/vllm/commit/37cecc8141e2dcba9dd534897a164d4baaf27229
• https://github.com/vllm-project/vllm/releases/tag/v0.32.0