CVE-2026-54234

HIGHCVSS 7.5/10EPSS 0.34%

Last modified

CVE-2026-54234 is a high-severity vulnerability rated 7.5/10 on the CVSS scale. 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. EPSS estimates a 0.34% chance of exploitation in the next 30 days.

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.

Metrics

CVSS 3.1
7.5/10

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H

EPSS Probability
0.34%

25.9th percentile

Probability of exploitation in the next 30 days. Learn more

Weakness Enumeration

Affected Software

VendorProductVersions
VllmVllm< 0.24.0

References

Timeline

Published
Last Modified
Status
Analyzed

Frequently Asked Questions

What is CVE-2026-54234?
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.
How severe is CVE-2026-54234?
CVE-2026-54234 has a CVSS score of 7.5/10 (HIGH severity). The EPSS model estimates a 0.34% probability of exploitation in the next 30 days.
How do I fix CVE-2026-54234?
Check the vendor references and advisories linked above for patched versions and mitigation guidance. You can also run a Strix scan to test if your systems are affected.

How Strix Helps

Related CVEs from 2026

Are you affected by CVE-2026-54234?

Run a free Strix scan to check your systems for this vulnerability.

Scan your code now

Source: NVD / NIST