CVE-2026-100654
Last modified
CVE-2026-100654 is a medium-severity vulnerability rated 6.5/10 on the CVSS scale. 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. EPSS estimates a 0.31% chance of exploitation in the next 30 days.
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).
Metrics
Weakness Enumeration
Affected Software
| Vendor | Product | Versions |
|---|---|---|
| Vllm | Vllm | < 0.29.0 |
References
- https://github.com/vllm-project/vllm/security/advisories/GHSA-v5gm-qgmv-gc6cMitigation, Vendor Advisory
Timeline
- Published
- Last Modified
- Status
- Analyzed
Frequently Asked Questions
What is CVE-2026-100654?
How severe is CVE-2026-100654?
How do I fix CVE-2026-100654?
How Strix Helps
- How Strix found a critical auth bypass in etcdStrix autonomously discovered a critical authentication bypass in etcd, later designated CVE-2026-33413.
- Autonomous PentestingAI agents that find and validate exploitable vulnerabilities like this one across your applications.
- PR ReviewsPentest every pull request so vulnerable code is caught before it ships to production.
- AI Penetration TestingHow AI-driven penetration testing continuously covers your attack surface.
Related CVEs from 2026
- CVE-2026-100649vLLM before 0.29.0 contains a resource-limit bypass vulnerab…5.3
- CVE-2026-10065A weakness has been identified in Shibby Tomato 1.28. This v…8.8
- CVE-2026-100650vLLM through 0.29.0 fetches and fully materializes remote or…6.5
- CVE-2026-100651vLLM before 0.29.0 fails to enforce decoder prompt-length va…6.5
- CVE-2026-100652vLLM versions 0.22.0 through 0.23.0 fail to validate stop_to…7.5
- CVE-2026-100653vLLM is an inference and serving engine for large language m…6.5
- CVE-2026-100655Rejected reason: This CVE ID has been rejected as a duplicat…
- CVE-2026-100656Rejected reason: This CVE ID has been rejected as a duplicat…
- CVE-2026-100657Rejected reason: This CVE ID has been rejected as a duplicat…
- CVE-2026-100658Rejected reason: This CVE ID has been rejected as a duplicat…
- CVE-2026-100659Netty's HTTP/3 codec (io.netty:netty-codec-http3) in version…6.5
- CVE-2026-10066A security vulnerability has been detected in Shibby Tomato …8.8
Are you affected by CVE-2026-100654?
Run a free Strix scan to check your systems for this vulnerability.
Scan your code nowSource: NVD / NIST
