vLLM before 0.29.0 Denial of Service via out-of-range stop_token_ids (CVE-2026-100654) | HOL Guard CVE

High · CVSS 7.1CVE-2026-100654

vLLM before 0.29.0 Denial of Service via out-of-range stop_token_ids

Answer in brief

CVE-2026-100654 records a High severity (CVSS 7.1) vulnerability in vLLM before 0.29.0 Denial of Service via out-of-range stop_token_ids. The current sources do not mark it as known exploited. The current feed maps vllm-project/vllm (generic). Check affected ranges and fixed versions before updating.

Analysis pending evidence review

HOL Guard separates source facts from reviewed analysis. See the methodology.

Published Sep 26, 2026Updated Sep 26, 2026Source checked Sep 26, 2026First seen by HOL Sep 26, 2026Material review Sep 26, 2026

Upstream Advisory

Key facts

Risk: High · CVSS 7.1
Exploitation: Not marked as known exploited
Affected software: 1 mapped package or product
Fix availability: Available

Why this deserves its current priority

CVSS is 7.1. The current sources do not mark it as known exploited. Treat this as a source-backed prioritization signal, not a statement about your environment.

Analysis status

Analysis pending evidence review

Factual feed record only; HOL analysis is not approved for indexing. Read the methodology.

Affected scope and exposure questions

The current feed maps vllm-project/vllm (generic). Check affected ranges and fixed versions before updating.

Mapped affected packages and fixed versions
Package	Affected range	Fixed version
vllm-project/vllmgeneric	>=0 <0.29.0	0.29.0

Recommended response

Check inventory. Check lockfiles and deployed manifests for vllm-project/vllm.
Review the reported fix. Update vllm-project/vllm to 0.29.0 if you use the affected versions. Test the change in a non-production environment first.

Evidence timeline and material changes

Published upstream
Sep 26, 2026
Evidence: source:cvelist:source_dates:source-dates:record
Source modified
Sep 26, 2026
Evidence: source:cvelist:source_dates:source-dates:record
First seen by HOL
Sep 26, 2026

Sources and claim methodology

Upstream source 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).

Quoted source text, attributed separately from HOL analysis.