CVE-2026-54234: vLLM denial of service through speculative decoding over gRPC Generate

Summary

CVE-2026-54234 affects vLLM before 0.24.0, an inference and serving engine for LLMs. A frontend-legal multi-request speculative decoding workload makes the rejection sampler emit a recovered token at the vocabulary size boundary, which becomes -1 and reaches the model's embedding and attention path, crashing the engine worker with a GPU device-side assertion. A remote client that can send requests through the public gRPC Generate and Abort endpoints can trigger this, aborting concurrent requests and causing a service-wide denial of service until the worker restarts.

Mitigation

Fixed in version 0.24.0.

Affected packages in the Exposure Registry

Matched by package name and ecosystem. Each entry shows whether the package delegates to a language model and how many tracked packages depend on it.

• vllmPyPILLM dependency since 2023-08-25 · 15 tracked dependents

Topics

• Inference infrastructure

Related items

• MediumCVE-2026-109311: vLLM audio upload heap overflow via FFmpeg IAMF parserSimilar attack · NVD/CVE Database
• MediumCVE-2026-108748: Quarkus LangChain4j memory exhaustion in chat-scopes WebSocketSimilar attack · NVD/CVE Database
• HighHermes Agent - Pre-Authentication Disk ConsumptionSimilar attack · Tenable Research Advisories
• MediumHermes Agent - Pre-Authentication Memory ExhaustionSimilar attack · Tenable Research Advisories
• MediumGHSA-v36g-jcw9-x7cw: Pydantic AI: Excessive resource use when local web fetching converts nested HTMLSimilar attack · GitHub Advisory Database