{"id":"CVE-2026-100654","title":"vLLM before 0.29.0 accepts user-controlled stop_token_ids on the OpenAI-compatible POST /v1/completions and POST /v1/chat/completions endpoints but validates only that the values are integers, not that each token id is within the model v…","summary":"vLLM before 0.29.0 accepts user-controlled stop_token_ids on the OpenAI-compatible POST /v1/completions and POST /v1/chat/completions endpoints but validates only that the values are integers, not that each token id is within the model v…","severity":"medium","cvss":6.5,"cvssVector":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H","cwe":["CWE-129","CWE-617"],"vendor":"vllm","product":"vllm","affected":["vllm < 0.29.0"],"patched":["vllm 0.29.0"],"published":"2026-09-26","updated":"2026-10-07","sourceUpdated":"2026-10-07T17:33:53.093","source":"NVD","sourceUrl":"https://nvd.nist.gov/vuln/detail/CVE-2026-100654","references":[{"url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-v5gm-qgmv-gc6c","label":"disclosure@vulncheck.com"},{"url":"https://www.vulncheck.com/advisories/vllm-before-0.29.0-denial-of-service-via-out-of-range-stop-token-ids","label":"disclosure@vulncheck.com"},{"url":"https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-100654.json"},{"url":"https://access.redhat.com/security/cve/CVE-2026-100654"},{"url":"https://bugzilla.redhat.com/show_bug.cgi?id=2541916"},{"url":"https://www.cve.org/CVERecord?id=CVE-2026-100654"},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-100654"}],"tags":["nvd","cve.org","csaf","vex","red-hat"],"epss":0.00314,"epssPercentile":0.24109,"ssvc":{"exploitation":"none","automatable":"no","technicalImpact":"partial","timestamp":"2026-09-30T14:51:29.949766Z"},"ingestedAt":"2026-09-26T13:33:20.917Z","slug":"CVE-2026-100654","body":"## Overview\n\nvLLM before 0.29.0 accepts user-controlled stop_token_ids on the OpenAI-compatible POST /v1/completions and POST /v1/chat/completions endpoints but validates only that the values are integers, not that each token id is within the model vocabulary/logits range. When min_tokens > 0, the stop token ids are used as logits indices to suppress stop tokens, so an out-of-range id reaches a CUDA indexing operation (index_put_) and triggers a device-side assertion. An authenticated API user can send a single malformed completion request that returns 500 Internal Server Error and puts EngineCore into a fatal state, causing subsequent requests to fail until the service is restarted (denial of service).\n\n## Affected\n\n- `vllm < 0.29.0`\n\n## Remediation\n\nUpgrade past the affected range:\n\n- `vllm 0.29.0`\n\n## Vendor advisories\n\n- **Red Hat VEX** · Moderate · 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 2026-10-01 · [vex](https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-100654.json)","depth":"sunlit","depthScore":36,"depthScoreParts":{"impact":35.8,"likelihood":0.1,"exploitation":0,"ransomware":0},"changes":[]}