vllm
Affected
- introduced
0, fixed0.29.0(ecosystem)
Fixed versions
0.29.0
Fix
- Fixed version
0.29.0pip install --upgrade vllm==0.29.0
Advisory
Also known as: CVE-2026-93840
The record has no summary.
vLLM before 0.29.0 validates allowed_token_ids against tokenizer length instead of model output logits width in SamplingParams._validate_allowed_token_ids(). Attackers can supply token IDs above the output vocabulary that pass validation, causing LogitBiasState to corrupt GPU logits state and allow concurrent requests to sample tokens outside their allowlists.
Medium Basis: CVSS v3 score 5.3, vector CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:L/A:N
0, fixed 0.29.0 (ecosystem)0.29.0
0.29.0pip install --upgrade vllm==0.29.0Fix commands are the export's own generic commands. Ojuju never runs them. Check them against your setup first.
Check your own computer: guard packages check
Source: PyPA Advisory Database · CC BY 4.0