# CVE-2026-114051: vLLM denial of service via uncaught exception in request validation

Source: https://junglewise.ai/threats/cve-2026-114051-vllm-before-0-31-0-skips-the-min-tokens-max-tokens-check-for  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-10-21. Updated: 2026-10-21  
License: CC BY 4.0 with a link back

CVE-2026-114051 · Severity: high · CVSS 7.5 · Published 2026-10-21

vLLM is a language model serving engine that processes AI inference requests. An unauthenticated attacker can send a single malicious request to the /inference/v1/generate endpoint that causes the engine's input thread to crash with an uncaught exception, halting all subsequent inference requests indefinitely. The service appears healthy to monitoring (health checks return success) but becomes completely unresponsive, requiring manual restart.

Technologies: [vLLM Project vllm](https://junglewise.ai/threats/technologies/vllm-project-vllm). Vendors: [vLLM Project](https://junglewise.ai/threats/vendors/vllm-project).

## Quick answers

**Is CVE-2026-114051 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-114051 being exploited in the wild.

**How severe is CVE-2026-114051?** High severity, with a CVSS score of 7.5, published 21 October 2026.

**Which products and versions are affected?** vLLM Project vllm before 0.31.0.

**Is there a fix for CVE-2026-114051?** Yes: fixed in 0.31.0. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.

**How is CVE-2026-114051 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-114051?** Check whether you run vLLM Project vllm at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## How to fix

- Fixed in 0.31.0

## Technical details

The vulnerability stems from a skipped min_tokens ≤ max_tokens validation check when max_tokens is null in /inference/v1/generate requests. The API process creates SamplingParams without validation, the EngineCore later rebuilds the params, validation fails with an uncaught VLLMValidationError on the decode line, and the exception is not caught by the surrounding try-except, killing the input thread. This is a denial-of-service via improper exception handling (CWE-248, CWE-754) requiring only network access.

## Affected products

- vLLM Project vLLM before 0.31.0

## Timeline

- 2026-10-06: disclosed
- 2026-10-20: advisory
- 2026-10-21: patched

## References

- [https://github.com/vllm-project/vllm/security/advisories/GHSA-4xqp-c3mv-qff7](https://github.com/vllm-project/vllm/security/advisories/GHSA-4xqp-c3mv-qff7)
- [https://www.vulncheck.com/advisories/vllm-before-0.31.0-denial-of-service-via-min-tokens-in-inference-v1-generate](https://www.vulncheck.com/advisories/vllm-before-0.31.0-denial-of-service-via-min-tokens-in-inference-v1-generate)

## Related threats

- [vLLM chat template resource exhaustion DoS](https://junglewise.ai/threats/cve-2026-115918-vllm-is-an-inference-and-serving-engine-for-large-language)
- [vLLM off-by-one error in tokenizer max_token_id validation](https://junglewise.ai/threats/cve-2026-115916-vllm-is-an-inference-and-serving-engine-for-large-language)
- [vLLM image format validation bypass in media processing](https://junglewise.ai/threats/cve-2026-115915-vllm-is-an-inference-and-serving-engine-for-large-language)
- [vLLM Phi-4 multimodal audio encoder GPU memory exhaustion](https://junglewise.ai/threats/cve-2026-115913-vllm-is-an-inference-and-serving-engine-for-large-language)
- [vLLM arbitrary code execution in shared-memory multimodal cache](https://junglewise.ai/threats/cve-2026-115912-vllm-is-an-inference-and-serving-engine-for-large-language)