Medium severity6.5NVD Advisory· Published Jun 22, 2026· Updated Jun 24, 2026
CVE-2026-54235
CVE-2026-54235
Description
vLLM is an inference and serving engine for large language models (LLMs). Prior to 0.23.1rc0, ll temperature validation gates use comparison operators (<, >), which silently evaluate to False for NaN and for positive Infinity in Python's IEEE 754 float semantics. Both values pass every guard and propagate to GPU sampling kernels, where they produce undefined behavior or CUDA errors that can crash the inference worker. This vulnerability is fixed in 0.23.1rc0.
AI Insight
LLM-synthesized narrative grounded in this CVE's description and references.
Affected packages
Versions sourced from the GitHub Security Advisory.
| Package | Affected versions | Patched versions |
|---|---|---|
vllmPyPI | >= 0.8.5, < 0.24.0 | 0.24.0 |
Affected products
8- osv-coords6 versionspkg:apk/chainguard/py3.12-vllm-cuda-12.4pkg:apk/chainguard/vllm-cuda-13.2pkg:apk/chainguard/py3.10-vllm-cuda-12.4pkg:apk/chainguard/tritonserver-backend-vllm-cuda-13.0pkg:apk/chainguard/vllm-openai-cuda-12.9pkg:apk/chainguard/vllm-openai-cuda-13.0
< 0.18.1-r5+ 5 more
- (no CPE)range: < 0.18.1-r5
- (no CPE)range: < 0.24.0-r0
- (no CPE)range: < 0.18.1-r5
- (no CPE)range: < 25.11-r12
- (no CPE)range: < 0.28.0-r0
- (no CPE)range: < 0.24.0-r1
Patches
Vulnerability mechanics
References
7- github.com/vllm-project/vllm/commit/d598d239737cfa37bcfcb98886ec3f3557fc7198nvdPatchWEB
- github.com/vllm-project/vllm/security/advisories/GHSA-7h4p-rffg-7823nvdExploitThird Party AdvisoryWEB
- github.com/advisories/GHSA-7h4p-rffg-7823ghsaADVISORY
- nvd.nist.gov/vuln/detail/CVE-2026-54235ghsaADVISORY
- github.com/pypa/advisory-database/tree/main/vulns/vllm/PYSEC-2026-3405.yamlghsaWEB
- github.com/vllm-project/vllm/pull/45116nvdIssue TrackingWEB
- pypi.org/project/vllmghsaWEB
News mentions
1- vLLM: Six CVEs Disclosed in 21 Hours — Critical Auth Bypass, Code Execution, and GPU Memory LeaksVypr Intelligence · Jun 17, 2026