{"id":"CVE-2026-94626","aliases":[],"url":"https://o3.security/vulnerability/CVE-2026-94626","summary":"vLLM through 0.29.0 Memory Exhaustion via Unvalidated NIXL tp_size","details":"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 disaggregated deployments to exhaust memory and trigger kernel OOM-kill of the decode worker process.","published":"2026-09-21T22:04:15.352Z","modified":"2026-09-22T11:45:54.050790298Z","cvss":null,"epss":null,"cisaKev":null,"exploitsKnown":null,"affectedPackages":[],"fix":{"url":"https://github.com/vllm-project/vllm/pull/51137","label":"vllm-project/vllm#51137"},"references":[{"type":"ADVISORY","url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/94xxx/CVE-2026-94626.json"},{"type":"PACKAGE","url":"https://github.com/vllm-project/vllm"},{"type":"ARTICLE","url":"https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/distributed/kv_transfer/kv_connector/utils.py#L569-L573"},{"type":"ARTICLE","url":"https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/distributed/kv_transfer/kv_connector/v1/nixl/metadata.py#L277"},{"type":"FIX","url":"https://github.com/vllm-project/vllm/pull/51137"},{"type":"ADVISORY","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-94626"},{"type":"ADVISORY","url":"https://www.vulncheck.com/advisories/vllm-through-0.29.0-memory-exhaustion-via-unvalidated-nixl-tp-size"}],"provenance":{"sources":["OSV.dev","FIRST.org (EPSS)"],"lastVerified":"2026-09-22T11:45:54.050790298Z"}}