CVE-2026-126397

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.

Severity
HIGH
CVSS v3
7.5
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H · source cna
CVSS v4
8.7
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N · source cna
Exploit chance (EPSS)
0.464%
chance of exploitation in the next 30 days
higher than 40.9% of scored CVEs · 2026-11-18
CISA KEV
Not listed
Public exploit code
None public
Exploit references (NVD and CNA): 0
Metasploit modules: 0
Nuclei templates: 0

Weakness

Timeline

  1. 2026-11-13Published by the CNA09:35 UTC

The timeline shows what already happened. To be told when this CVE changes next, watch it from the API.

Affected products

Products listed as vulnerable. Upgrade to the version under "Fixed in", or a later one.

ProductVulnerable versionsFixed in
vllm vllmfrom 0.10.1 before 0.31.0fixed in 0.31.00.31.0
vllm-project vllmfrom 0.10.1 before 0.31.0fixed in 0.31.00.31.0

Products and versions as NVD and the vendor that published the CVE list them.

References

Get more CVE data from the API.

Free without a key or signup. Pro adds monitoring and higher limits: pricing.

Sources