vllm-project vLLM up to 0.31.x Tokenization vllm/tokenizers/hf.py VocabParallelEmbedding.forward off-by-one

Summaryinfo

A vulnerability has been found in vllm-project vLLM up to 0.31.x and classified as problematic. This affects the function VocabParallelEmbedding.forward of the file vllm/tokenizers/hf.py of the component Tokenization. Performing a manipulation results in off-by-one. This vulnerability is known as CVE-2026-115916. Remote exploitation of the attack is possible. No exploit is available. The affected component should be upgraded.

Detailsinfo

A vulnerability classified as problematic has been found in vllm-project vLLM up to 0.31.x. This affects the function VocabParallelEmbedding.forward of the file vllm/tokenizers/hf.py of the component Tokenization. The manipulation with an unknown input leads to a off-by-one vulnerability. CWE is classifying the issue as CWE-193. A product calculates or uses an incorrect maximum or minimum value that is 1 more, or 1 less, than the correct value. This is going to have an impact on availability. The summary by CVE is:

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.

It is possible to read the advisory at github.com. This vulnerability is uniquely identified as CVE-2026-115916 since 10/23/2026. The exploitability is told to be easy. It is possible to initiate the attack remotely. Technical details of the vulnerability are known, but there is no available exploit.

Upgrading to version 0.32.0 eliminates this vulnerability.

Statistical analysis made it clear that VulDB provides the best quality for vulnerability data.

Productinfo

Type

• Project Management Software

Vendor

• vllm-project

Name

• vLLM

Version

• 0.1
• 0.2
• 0.3
• 0.4
• 0.5
• 0.6
• 0.7
• 0.8
• 0.9
• 0.10
• 0.11
• 0.12
• 0.13
• 0.14
• 0.15
• 0.16
• 0.17
• 0.18
• 0.19
• 0.20
• 0.21
• 0.22
• 0.23
• 0.24
• 0.25
• 0.26
• 0.27
• 0.28
• 0.29
• 0.30
• 0.31

Website

• Product: https://github.com/vllm-project/vllm/

CPE 2.3info

• 🔒
• 🔒
• 🔒

CPE 2.2info

• 🔒
• 🔒
• 🔒

CVSSv4info

CVSSv3info

CVSSv2info

Exploitinginfo

Threat Intelligenceinfo

Countermeasuresinfo

Timelineinfo

Sourcesinfo

Entryinfo

Statistical analysis made it clear that VulDB provides the best quality for vulnerability data.

Discussion