CVE Tools

CVE-2026-100654

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

Published: Sep 26, 2026Updated: Sep 26, 2026 Sources: CVE List NVDCWE-129

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).

No summary for this CVE yet.

CVSS Vector Breakdown

AV:NAC:LPR:LUI:NS:UC:NI:NA:H
Exploitability
AV:NAttack Vector
Network
AC:LAttack Complexity
Low
PR:LPrivileges Required
Low
UI:NUser Interaction
None
Scope
S:UScope
Unchanged
Impact
C:NConfidentiality
None
I:NIntegrity
None
A:HAvailability
High

Weaknesses

Affected Products

vllm-project
oss-projectaka vllm, vllm-project/vllm

Exploitability

No known exploits, KEV entries, or remediation guidance available for this vulnerability yet.

References

Unlock Complete Vulnerability Intelligence

Get the full picture for CVE-2026-100654 and every CVE in our database. Create a free account — no credit card required.

Create Free Account
Plain-language analysis
Impact assessment and exploitation scenario in plain English
Attack graph visualization
Interactive attack path and kill chain mapping
Exploit details & PoC links
ExploitDB, Metasploit, GitHub PoCs with direct links
Nuclei scanner templates
Ready-to-use vulnerability scanner templates
Full remediation guide
Patch instructions, workarounds, and compliance impact
Interactive AI chat
Ask questions about this vulnerability in natural language
Related vulnerabilities
Semantically similar CVEs and attack patterns
REST API & MCP access
Integrate vulnerability data into your workflows

We use analytics cookies to see which pages and articles actually help people. Decline and none of them run — the site works the same. What we store