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.
PUBLISHED5.2CWE-1287
vLLM: temperature=NaN and temperature=Infinity bypass validation and propagate to GPU kernels
Problem type
Affected products
vllm-project
vllm
< 0.23.1rc0 - AFFECTED
References
https://github.com/vllm-project/vllm/security/advisories/GHSA-7h4p-rffg-7823
https://github.com/vllm-project/vllm/security/advisories/GHSA-7h4p-rffg-7823
https://github.com/vllm-project/vllm/pull/45116
https://github.com/vllm-project/vllm/pull/45116
https://github.com/vllm-project/vllm/commit/d598d239737cfa37bcfcb98886ec3f3557fc7198
https://github.com/vllm-project/vllm/commit/d598d239737cfa37bcfcb98886ec3f3557fc7198
GitHub Security Advisories
GHSA-7h4p-rffg-7823
vLLM: temperature=NaN and temperature=Infinity bypass validation and propagate to GPU kernels
https://github.com/advisories/GHSA-7h4p-rffg-7823Summary
All 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. Note: -Infinity is correctly caught.
Root Cause
sampling_params.py:384:
if 0 < self.temperature < _MAX_TEMP: # NaN → False; +Inf → False
sampling_params.py:462:
if self.temperature < 0.0: # NaN → False; +Inf → False
raise VLLMValidationError(...)
No math.isnan() or math.isinf() check exists anywhere in sampling_params.py.
Python semantics (verified): float('nan') < 0.0 → False, float('inf') < 0.0 → False.
Impact
Crash of inference worker on GPU kernel execution with NaN/Inf softmax input, degrading service for all concurrent users.
Remediation
Add math.isfinite(self.temperature) check in _verify_args(). Reject non-finite float values with a 400 error.
Fix
A fix for this vulnerability was merged here: https://github.com/vllm-project/vllm/pull/45116
JSON source
https://cveawg.mitre.org/api/cve/CVE-2026-54235Click to expand
{
"dataType": "CVE_RECORD",
"dataVersion": "5.2",
"cveMetadata": {
"cveId": "CVE-2026-54235",
"assignerOrgId": "a0819718-46f1-4df5-94e2-005712e83aaa",
"assignerShortName": "GitHub_M",
"dateUpdated": "2026-06-22T21:59:02.710Z",
"dateReserved": "2026-06-12T16:25:43.084Z",
"datePublished": "2026-06-22T21:59:02.710Z",
"state": "PUBLISHED"
},
"containers": {
"cna": {
"providerMetadata": {
"orgId": "a0819718-46f1-4df5-94e2-005712e83aaa",
"shortName": "GitHub_M",
"dateUpdated": "2026-06-22T21:59:02.710Z"
},
"title": "vLLM: temperature=NaN and temperature=Infinity bypass validation and propagate to GPU kernels",
"descriptions": [
{
"lang": "en",
"value": "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."
}
],
"affected": [
{
"vendor": "vllm-project",
"product": "vllm",
"versions": [
{
"version": "< 0.23.1rc0",
"status": "affected"
}
]
}
],
"problemTypes": [
{
"descriptions": [
{
"lang": "en",
"description": "CWE-1287: Improper Validation of Specified Type of Input",
"cweId": "CWE-1287",
"type": "CWE"
}
]
}
],
"references": [
{
"url": "https://github.com/vllm-project/vllm/security/advisories/GHSA-7h4p-rffg-7823",
"name": "https://github.com/vllm-project/vllm/security/advisories/GHSA-7h4p-rffg-7823",
"tags": [
"x_refsource_CONFIRM"
]
},
{
"url": "https://github.com/vllm-project/vllm/pull/45116",
"name": "https://github.com/vllm-project/vllm/pull/45116",
"tags": [
"x_refsource_MISC"
]
},
{
"url": "https://github.com/vllm-project/vllm/commit/d598d239737cfa37bcfcb98886ec3f3557fc7198",
"name": "https://github.com/vllm-project/vllm/commit/d598d239737cfa37bcfcb98886ec3f3557fc7198",
"tags": [
"x_refsource_MISC"
]
}
],
"metrics": [
{}
]
}
}
}