
Signal is active with 1 mentions in latest observed window
Recommended action window: Monitor and triage in normal cycle
NVD description
vLLM from 0.10.1 before 0.31.0 forwards a request-controlled mm_processor_kwargs.max_patches value to the Hugging Face processor for Cohere2 Vision models without a server-side upper bound, causing quadratic CPU work during multimodal preprocessing. Remote clients can send small image chat-completion requests with large, varying max_patches values to bypass the LRU cache, exhaust preprocessing workers, and deny service to other clients.
Priority
LOW
Exploitation
NONE
PoC
NONE
Patch
NONE
Momentum
NONE

🚨 HIGH - Resource Exhaustion DoS in vLLM Cohere2 Vision Models (CVE-2026-126397) A high-severity denial-of-service vulnerability has been disclosed in vLLM affecting multimodal serving for Cohere2 Vision models, allowing remote clients to exhaust preprocessing workers. 👉 Affected / Remediation: Affects vLLM versions 0.10.1 up to 0.30.x. Upgrade immediately to vLLM 0.31.0 or later.