CVE
vLLM denial of service in multimodal image preprocessing
Affects vLLM 0.10.1 before 0.31.0 when serving Cohere2 Vision models through the Python OpenAI-compatible frontend. Fixed in 0.31.0.
Written by anthropic/claude-haiku-5.5 from the sources below. Verify against them.
- Exploited
- No known exploitation
- Fix
- Fixed in 0.31.0
- Who is exposed
- Default setups
Do first: Check every install: default setups are affected.
- Published
- Last modified
- Whether it needs action now or is a reference record, by fixed rules, with why. Last checked is when cve.now last re-read it.What do these mean? →
- Reference
- Last checked
- A 0-100 score for how much a CVE matters across the industry: exploitation, exposure, how widely the software is used, severity and whether a fix exists.What do these mean? →
- 31/100 · provisional
- How many real-world setups can hit the bug, from the vendor's advisory and CISA data. It does not know your setup.What do these mean? →
- Default setups
- Whether working attack code is public (Exploit-DB, Metasploit, Nuclei or GitHub). Public code makes attacks easier.What do these mean? →
- None on Exploit-DB, Metasploit or GitHub
Summary
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.
Who should check
Teams running vLLM from 0.10.1 before 0.31.0 that serve Cohere2 Vision models through the OpenAI-compatible server and accept image chat requests from clients.
Not fixed by: Explicitly enabling --trust-request-mm-kwargs restores trusted override behavior.
cve.now Index
- Exploitation?/40Unknown. No CISA SSVC exploitation rating yet
- Exposure22/25Default setups, reachable over the network without a login
- Reach?/20Unknown. No tracked technology or package matched
- Severity8.7/10CVSS 8.7
- No patch0/5No signal
What would change the score
- Would score about 71 if added to CISA KEV
- Would score about 66 if CISA's SSVC rates exploitation active
What to do
Check every install: default setups are affected.
Upgrade to a fixed version: 0.31.0.
Check your systems with an agent
Paste this into a coding or ops agent. It checks your systems read-only and reports a table per asset.
Check whether my systems are affected by CVE-2026-126397 (vLLM 0.10.1 before 0.31.0 Cohere2 Vision DoS via max_patches), using the cve.now data below.
Rules: work read-only. Do not run exploit code, proof-of-concept scripts or scanners that send attack payloads, and do not change any configuration, package, image or deployment. Report findings only.
What cve.now has for CVE-2026-126397 (entry last updated 2026-11-13):
Affected:
- vllm-project vllm: from 0.10.1, fixed in 0.31.0
Fixed versions: 0.31.0.
Conditions (all must be true for the bug to be hit):
- vLLM is serving a Cohere2 Vision model through the Python OpenAI-compatible frontend. (source: https://github.com/vllm-project/vllm/security/advisories/GHSA-9g3c-x747-c6wr)
- The caller can submit an image chat request. (source: https://github.com/vllm-project/vllm/security/advisories/GHSA-9g3c-x747-c6wr)
Setup: the bug is reachable in a default setup; no special feature or setting is needed.
Attacker needs: network access, no login, no user action.
CISA KEV: CVE-2026-126397 is not listed as of cve.now's last read.
This KEV status is for CVE-2026-126397 only. If you also check another CVE, look up its own KEV status; never assume it from this one.
How to check:
- Installed software: check what is actually running, not what a repo declares. Read the version from each host, container image and Kubernetes node (the binary's version output, the OS package database, image layers), and check whether the conditions above hold there. Distribution packages may carry a backported fix under an older version number; confirm against the distribution's own advisory.
Verify against the linked advisories before concluding, and cite the date you read each one:
- https://github.com/vllm-project/vllm/security/advisories/GHSA-9g3c-x747-c6wr
- https://github.com/vllm-project/vllm/commit/f6326f53bda46898a331c2d24500332c285d9a2b
- https://www.vulncheck.com/advisories/vllm-0.10.1-before-0.31.0-cohere2-vision-dos-via-max-patches
Do not conclude an asset is unaffected unless you checked it and the evidence covers both its version and the conditions above. If you could not check something, mark it unknown and say what is missing.
Report a table with one row per asset and these columns: Asset | Version or build | Status (affected, fixed, not applicable, unknown) | Compensating controls | Exposure (internet-facing, internal, none) | Evidence | Next action.
Entry: https://cve.now/CVE-2026-126397
For follow-up questions, use the cve.now MCP server at https://mcp.cve.now/mcp (no authentication), for example execute with name cve.get and arguments {"id": "CVE-2026-126397"}.Timeline
- CVE published
- First seen by cve.now
Affected versions
- vllm-project vllm
From 0.10.1 · fixed in 0.31.0
Evidence 5
Vendor advisory · 1
Establishes: Affected and fixed versions.
- Advisory
CVE record · 2
Establishes: The identifier and description.
- Record
- Record
Corrections
None so far.
Something wrong on this page? Report an error. Every fix is listed here.
Questions about CVE-2026-126397
Is CVE-2026-126397 being exploited?
It is not in CISA's Known Exploited Vulnerabilities catalog, and CISA's SSVC assessment does not mark exploitation as active. That does not mean it has not been attacked: vendors and researchers can report attacks before CISA lists them. No public exploit code is tracked.Source: cve.now feed data
What does this mean?
A remote caller who submits an image chat request can make vLLM consume disproportionate CPU time in multimodal preprocessing, delaying or denying image requests from other clients. The demonstrated impact is limited to availability.Source: github.com
Which versions are affected by CVE-2026-126397?
vllm: from 0.10.1 before 0.31.0.Source: cve.now feed data
Am I affected, and how do I check?
Versions from 0.10.1 up to but not including 0.31.0 are in the affected range. The setup involves serving a Cohere2 Vision model through the Python OpenAI-compatible frontend and accepting image chat requests. Check the installed vLLM version and the model being served.Source: github.com
How do I fix CVE-2026-126397?
Upgrade vLLM to 0.31.0 or later.Source: github.com