{"id":"CVE-2026-94626","title":"vLLM through 0.29.0 fails to validate the tp_size parameter in kv_transfer_params on OpenAI-compatible completion endpoints, allowing attackers to allocate unbounded memory","summary":"vLLM through 0.29.0 fails to validate the tp_size parameter in kv_transfer_params on OpenAI-compatible completion endpoints, allowing attackers to allocate unbounded memory. Attackers can supply arbitrary tp_size values in prefill/decode…","severity":"high","cvss":7.5,"cvssVector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","cwe":["CWE-789","CWE-770"],"vendor":"vllm-project","product":"vllm","affected":["vllm <= 0.29.0"],"published":"2026-09-21","updated":"2026-09-24","sourceUpdated":"2026-09-24T23:19:22.817","source":"NVD","sourceUrl":"https://nvd.nist.gov/vuln/detail/CVE-2026-94626","references":[{"url":"https://github.com/vllm-project/vllm","label":"disclosure@vulncheck.com"},{"url":"https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/distributed/kv_transfer/kv_connector/utils.py#L569-L573","label":"disclosure@vulncheck.com"},{"url":"https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/distributed/kv_transfer/kv_connector/v1/nixl/metadata.py#L277","label":"disclosure@vulncheck.com"},{"url":"https://github.com/vllm-project/vllm/pull/51137","label":"disclosure@vulncheck.com"},{"url":"https://www.vulncheck.com/advisories/vllm-through-0.29.0-memory-exhaustion-via-unvalidated-nixl-tp-size","label":"disclosure@vulncheck.com"},{"url":"https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-94626.json"},{"url":"https://access.redhat.com/security/cve/CVE-2026-94626"},{"url":"https://bugzilla.redhat.com/show_bug.cgi?id=2538419"},{"url":"https://www.cve.org/CVERecord?id=CVE-2026-94626"},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-94626"}],"tags":["nvd","cve.org","csaf","vex","red-hat"],"ssvc":{"exploitation":"none","automatable":"yes","technicalImpact":"partial","timestamp":"2026-09-24T22:54:32.814414Z"},"epss":0.0063,"epssPercentile":0.47907,"ingestedAt":"2026-09-21T22:54:37.974Z","slug":"CVE-2026-94626","body":"## Overview\n\nvLLM through 0.29.0 fails to validate the tp_size parameter in kv_transfer_params on OpenAI-compatible completion endpoints, allowing attackers to allocate unbounded memory. Attackers can supply arbitrary tp_size values in prefill/decode disaggregated deployments to exhaust memory and trigger kernel OOM-kill of the decode worker process.\n\n## Remediation\n\nRefer to the linked advisories for vendor-supplied fixes and affected version ranges.\n\n## Vendor advisories\n\n- **Red Hat VEX** · Important · affected: Red Hat AI Inference Server, Red Hat Enterprise Linux AI (RHEL AI) 3, Red Hat OpenShift AI (RHOAI) · no fix planned: Red Hat AI Inference Server, Red Hat Enterprise Linux AI (RHEL AI) 3, Red Hat OpenShift AI (RHOAI) · updated 2026-09-22 · [vex](https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-94626.json)","depth":"twilight","depthScore":41,"depthScoreParts":{"impact":41.3,"likelihood":0.1,"exploitation":0,"ransomware":0},"changes":[]}