CVE-2026-100654
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).
Does this matter?
High impact if exploited, but EPSS currently rates exploitation as unlikely (0.31%). Schedule it in the normal patch cycle and watch for a rise in EPSS or a public exploit.
Description
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 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).
- CVSS 4.0
- 7.1 HIGHCVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/M
- EPSS
- 0.31% probability · 24th percentile
- Public exploits
- None in Exploit-DB
- CISA KEV
- Not listed
- Weakness
- CWE-129
- Affected
- vllm/vllm
- Source
- disclosure@vulncheck.com
EPSS moved from 0.3% on 27 September 2026 to 0.3% on 26 October 2026.
References
- https://github.com/vllm-project/vllm/security/advisories/GHSA-v5gm-qgmv-gc6cMitigation, Vendor Advisory
- https://www.vulncheck.com/advisories/vllm-before-0.29.0-denial-of-service-via-out-of-range-stop-token-idsThird Party Advisory
Source: NVD record, EPSS from FIRST.org, KEV from CISA, exploits from Exploit-DB. Refreshed daily. Download this record as JSON.