CVE-2026-54234
Vllm vulnerability
- Published6 Jul 2026
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.
CVSS 3.1 breakdown
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:HAffected products
Remediation
References
Related CVEs
Recent high-severity vulnerabilities sharing a weakness type.
Frequently asked
Is CVE-2026-54234 being actively exploited?
CVE-2026-54234 is not in the CISA KEV catalog. A public exploit reference exists, so treat it as exploitable. Its EPSS exploit probability is 0.62%.
How severe is CVE-2026-54234?
CVE-2026-54234 has a CVSS 3.1 base score of 7.5, rated high. EPSS estimates a 0.62% probability of exploitation.
Which products are affected by CVE-2026-54234?
Affected products include Vllm.
How do I fix CVE-2026-54234?
Upgrade affected software to a fixed version, for example Vllm 0.24.0 or later, then confirm no affected versions remain in your inventory.
How Secuno handles this
Secuno correlates vulnerabilities like CVE-2026-54234 against the real software inventory on every managed device, every day. Rather than a static severity, each match is scored in context: exploit availability, EPSS, the asset's exposure, and business criticality decide where it lands in the queue. If an affected version is present anywhere across your portfolio, it surfaces automatically with a prioritised remediation path.
Data from the NVD, CISA KEV catalog and FIRST EPSS. CVSS 3.1 base score shown where available.