vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. Prior to 0.24.0, a frontend-legal multi-request speculative decoding workload can cause the rejection sampler to produce a recovered token equal to the model vocabulary size boundary value, which is then converted to negative one when the engine selects the next live token for a request and is written back into the drafter's input ids; that out-of-vocabulary value is later consumed by the model's embedding and attention path and crashes the engine worker with a GPU device-side assertion. The same triggering request sequence is reachable through the public gRPC Generate and Abort endpoints, so a remote client that can send generation requests can crash the shared engine worker, aborting concurrent requests and causing a service-wide denial of service for other clients of the deployment until the worker is restarted. This issue is fixed in version 0.24.0.
Fix
Patch, vendor advisory and mitigation references on NVD 4
- https://github.com/vllm-project/vllm/commit/8a5cf1ccd65e8ac7635c402c1ec0b08988bc26ca Patch
- https://github.com/vllm-project/vllm/pull/44744 Issue Tracking, Patch
- https://github.com/vllm-project/vllm/security/advisories/GHSA-8wr5-jm2h-8r4f Vendor Advisory, Exploit
- https://github.com/vllm-project/vllm/security/advisories/GHSA-8wr5-jm2h-8r4f Vendor Advisory, Exploit
Every line above is quoted from NVD, CISA, Microsoft, the vendor's own advisory or DISA; nothing is generated.
Change history 1
- First seenCVSS 7.5 · vllmNVD modified 2026-10-25 20:10
Affected (CPE criteria) 1
cpe:2.3:a:vllm:vllm:*:*:*:*:*:*:*:*
References 4
- https://github.com/vllm-project/vllm/commit/8a5cf1ccd65e8ac7635c402c1ec0b08988bc26ca Patch
- https://github.com/vllm-project/vllm/pull/44744 Issue Tracking, Patch
- https://github.com/vllm-project/vllm/security/advisories/GHSA-8wr5-jm2h-8r4f Vendor Advisory, Exploit
- https://github.com/vllm-project/vllm/security/advisories/GHSA-8wr5-jm2h-8r4f Vendor Advisory, Exploit