Vulnerability Description
vLLM from 0.8.5 before 0.31.0 has a denial of service flaw. LogprobsTensors.empty_cpu() in vllm/v1/outputs.py allocates an uninitialized int32 tensor, and the V1 model runner leaves prompt positions reported as computed by a KV connector unwritten. A client sharing the engine can send prompt_logprobs (or echo with logprobs) against a connector-cached prefix, making tokenizer decode raise OverflowError, which crashes EngineCore and drops all co-tenant requests.
CVSS Score
HIGH
Affected Products
| Vendor | Product | Versions |
|---|---|---|
| Vllm | Vllm | >= 0.8.5, < 0.31.0 |
Related Weaknesses (CWE)
References
- https://github.com/vllm-project/vllm/security/advisories/GHSA-rhcx-5729-88vgMitigationVendor Advisory
- https://www.vulncheck.com/advisories/vllm-from-0.8.5-before-0.31.0-cross-tenant-Third Party Advisory
FAQ
What is CVE-2026-116180?
CVE-2026-116180 is a vulnerability with a CVSS score of 7.7 (HIGH). vLLM from 0.8.5 before 0.31.0 has a denial of service flaw. LogprobsTensors.empty_cpu() in vllm/v1/outputs.py allocates an uninitialized int32 tensor, and the V1 model runner leaves prompt positions r...
How severe is CVE-2026-116180?
CVE-2026-116180 has been rated HIGH with a CVSS base score of 7.7/10. Review the CVSS metrics above for detailed severity breakdown.
Is there a patch for CVE-2026-116180?
Check the references section above for vendor advisories and patch information. Affected products include: Vllm Vllm.