CVE-2026-54234
vLLM: Remote DoS in vLLM via Invalid Recovered Token Reinjection
Using vllm by vllm-project?
Get instant alerts when new vulnerabilities affect vllm - before attackers find them.
Description
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.
Remediation
No guide yetNo remediation guidance has been written for this CVE yet. What the record does show:
References (3)
SSVC — Action Prioritization
SSVC signals action urgency (what to patch first), separate from CVSS which measures severity.
CVSS
- Attack Vector
- NETWORK
- Complexity
- LOW
- Privileges
- NONE
- User Interaction
- NONE
- Scope
- UNCHANGED
- Confidentiality
- NONE
- Integrity
- NONE
- Availability
- HIGH
Quick Info
Get alerted
Stop reading CVE feeds looking for your own stack
Track the products you run and VulnTracker tells you when a new CVE affects them — with what is known about fixing it, and nothing about the rest.