vLLM is an inference and serving engine for large language models. From 0.11.1 until 0.32.0, the DeepSeek-OCR, DeepSeek-OCR-2, and Unlimited-OCR image path allows a remote client that can reach the OpenAI-compatible API to submit a legal image whose normalized pixels sum to zero when that image is alone in its per-step encoder batch. _parse_and_validate_image_input returns no embedding, embed_multimodal propagates that value, and sanity_check_mm_encoder_outputs triggers a fatal assertion that shuts down the engine for all users. This issue is fixed in version 0.32.0.
CVE-2026-115914
MEDIUMPre-NVD 6.56.5Trending — 5 sources updated this week
EchelonGraph scoreHIGH confidence
“—” beside the score: it is equal to the CVSS score beside it.
Score 6.5 from GitHub Security Advisory (severity: MODERATE) published 2026-10-28. A secondary CVSS source baseline 6.5; sources differ by 0.0.
Triggered by: GitHub Security Advisory CVSS
Sources: EPSS, GitHub Security Advisory, a secondary CVSS source
Also known asEUVD-2026-110032 (ENISA EUVD) · GHSA-gq5q-89c7-vhvx (GitHub Security Advisories)Source: ENISA EUVD
6.5EG
EchelonGraph verdictMonitorRemediate within your standard update timelines (the SSVC decision on this page).
- Lower severity and no public exploit yet
CISA-KEV: Not listedEPSS PROB: 0.4%CVSS: 6.5Exploit: None knownExposed services: Not assessed
A fix is available — apply it.
- CVSS v3
- 6.5
- EG Score
- 6.5MEDIUMhigh confidence
- EG Risk
- 29EG Risk 29/100
EG Risk is EchelonGraph's 0–100 priority score. It fuses impact severity (EG Severity, the CVSS base score, never the KEV-floored or EPSS-lifted EG Score) with real-world exploitation and automatability, under weights EchelonGraph set by hand, so you can rank equal-severity CVEs and fix the most dangerous first. Higher = act sooner.
How it’s computed (value × EchelonGraph’s weight)Severity65% × 45%Impact severity (EG Severity): CVSS 6.5 of 10. Never the KEV-floored or EPSS-lifted EG Score.Exploitation0.4% × 40%FIRST EPSS: the 30-day exploitation probability when EchelonGraph scored this CVE.Automatability0% × 15%CISA's SSVC Automatable value: no (Vulnrichment).SSVC (CISA's decision table, applied by EchelonGraph): Track at every mission impact level.Exploitation none (CISA Vulnrichment) · Automatable no (CISA Vulnrichment) · Technical impact partial (CISA Vulnrichment). All three are CISA's SSVC values (Vulnrichment). EchelonGraph applied CISA's decision table to them; CISA publishes SSVC inputs and the table, not a decision for each CVE.Action: A fix is available. Apply it within your standard update timelines. - EPSS PROB
- 0.4%
- EPSS %ILE
- 33rd
- KEV
- Not listed
What the EG score is made of (scorer v4.0: the three questions the 0–10 score mixes, published apart · methodology)
The EG score (6.5): severity, escalated by confirmed exploitation. This CVE's 6.5 is its impact severity (EG Severity), not escalated.
- EG Severity
- 6.5 MEDIUMCVSS-B, impact only: no KEV listing, EPSS or EG-KEV tier raises it.GitHub Security Advisory CVSS
- EG Likelihood
- 0.4% within 30 daysSource: FIRST EPSS (model v2026.06.15, score date 2026-10-31). FIRST's exploitation probability, not an EchelonGraph model.
- Confirmed exploitation
- None on record (CISA KEV, EG-KEV Tier 1)Reported apart from the likelihood: an observed fact, not a probability.
- EG Priority
- SSVC Track at medium mission impactTrack at every mission impact level.CISA BOD 26-04: 60 days if the affected system is publicly exposed; Fix on system upgrade if it is not.Each input and its source, and whom BOD 26-04 binds: the SSVC and BOD 26-04 blocks below.
SSVC (CISA's decision table, applied by EchelonGraph)Track at every mission impact level.
A fix is available. Apply it within your standard update timelines.
Exploitation none (CISA Vulnrichment) · Automatable no (CISA Vulnrichment) · Technical impact partial (CISA Vulnrichment). All three are CISA's SSVC values (Vulnrichment). EchelonGraph applied CISA's decision table to them; CISA publishes SSVC inputs and the table, not a decision for each CVE. Mission impact is CISA's Mission & Well-being decision point, and only you can judge it. CISA's table makes it high when Mission Prevalence is Essential — the vulnerable component "directly provides capabilities that constitute at least one MEF for at least one entity" (MEF: mission essential function) — or when Public Well-Being Impact is Irreversible: "multiple fatalities are likely", the cyber-physical system "is likely lost or destroyed", "extreme or serious externalities" are imposed on other parties, or social systems such as elections or the financial grid "are destabilized and potentially collapse". CISA's decision table
BOD 26-04 (CISA's remediation timeline, Table 1)60 days if the affected system is publicly exposed; Fix on system upgrade if it is not.
The score's inputs: In CISA KEV no (CISA's KEV catalog) · Automatable no (CISA Vulnrichment) · Technical impact partial (CISA Vulnrichment). Whether your system is publicly exposed is yours to answer. BOD 26-04 binds US Federal Civilian Executive Branch agencies; for anyone else it is CISA's published timeline, a reference rather than an obligation. The directive · how EchelonGraph applies it
Published
October 26, 2026
Last Modified
October 30, 2026
Advisory Details (4)
Auto-updated Oct 26, 2026Patch available. Sources: github, github_pr, github_commit, github_release.
GitHub release Patch Available
v0.32.0
Patch available: vllm-project/vllm v0.32.0
https://github.com/vllm-project/vllm/releases/tag/v0.32.0GitHub commit
commit c15672cfb491 (vllm-project/vllm)
Fix landed in vllm-project/vllm commit c15672cfb491 — awaiting tagged release
https://github.com/vllm-project/vllm/commit/c15672cfb49152ebf65cfd188e71b84593ec5686GitHub pull request
[Security] Accept zero-sum DeepSeek-OCR pixel tensors
Fix merged in vllm-project/vllm PR #59417 on 2026-10-02 — awaiting tagged release
https://github.com/vllm-project/vllm/pull/59417GitHub advisoryMentions a patch (unverified)
DeepSeek-OCR family: an image whose normalized pixels sum to exactly 0 makes embed_multimodal return None, tripping the encoder-output sanity assert — one unauthenticated request kills the whole vLLM engine (default config) · Advisory · vllm-project/vllm · GitHub
https://github.com/vllm-project/vllm/security/advisories/GHSA-gq5q-89c7-vhvxVendor Advisories for CVE-2026-115914(1)
These vendors published their own advisory mentioning this CVE — often with vendor-specific remediation steps + affected product lists not in NVD.
Patch Availability(1)
| Vendor / Ecosystem | Fix (by version range) / Patch | Released | Source |
|---|---|---|---|
| pip | vllm @ 0.32.0 (fixes >= 0.11.1, < 0.32.0) | — | ghsa |
Patches are aggregated from vendor advisories (Red Hat, Microsoft, Cisco, GitHub) and package ecosystems (OSV, GHSA). Multiple rows for the same upstream release have been deduplicated.
Affected Packages
(1 across 1 ecosystem)
PyPI(1)
| Package | Vulnerable range | Fix by version range | Dependents |
|---|---|---|---|
| vllm | 0.11.1 ... 0.31.0 (32 versions) |
| — |
Weakness Classification(1)
MITRE Common Weakness Enumeration — the root-cause categories this CVE belongs to.
Data Freshness Timeline
(refreshed 14× in last 7d / 14× in last 30d)
Each row is a time one of our pipelines fetched, updated or re-scored this CVE. For example, "NVD update" means NVD published or revised its analysis for this CVE; "MITRE cvelistV5" means we ingested or refreshed it from the CNA feed. An arrow marks a change the record states ("CVSS v3 → 9.1"); EPSS and GHSA rows show the values recorded at that time, not what changed (EPSS values are rounded; hover one for the exact figures). Most recent first.
- 2026-11-01 07:59 UTCGHSA enrichmentGHSA-gq5q-89c7-vhvx · GitHub severity MODERATE · GitHub CVSS 6.5
- 2026-11-01 07:59 UTCEG score recompute
- 2026-10-31 16:32 UTCEPSS rescoreEPSS 0.4% · 33rd percentile
- 2026-10-30 16:33 UTCEPSS rescoreEPSS 0.4% · 33rd percentile
- 2026-10-30 15:26 UTCEG score recompute
- 2026-10-30 14:35 UTCEG score recompute
- 2026-10-29 16:32 UTCEPSS rescoreEPSS 0.4% · 33rd percentile
- 2026-10-28 16:31 UTCEPSS rescoreEPSS 0.4% · 32nd percentile
- 2026-10-28 15:46 UTCEG score recompute
- 2026-10-27 16:31 UTCEPSS rescoreEPSS 0.4% · 32nd percentile
- 2026-10-26 19:07 UTCEG score recompute
- 2026-10-26 18:01 UTCENISA EUVD ID recordedEUVD-2026-110032 · Source: ENISA EUVD
- 2026-10-26 17:23 UTCEG score recomputeFirst assessed: 6.5 MEDIUM
- 2026-10-26 17:20 UTCMITRE cvelistV5first tracked
Related CVEs(same product + same CWE)
Same product
10 shownPyPI:vllm
Frequently asked(5)
What is CVE-2026-115914?
CVE-2026-115914 is a medium-severity vulnerability; its CVE record was published on October 26, 2026. vLLM is an inference and serving engine for large language models. From 0.11.1 until 0.32.0, the DeepSeek-OCR, DeepSeek-OCR-2, and Unlimited-OCR image path allows a remote client that can reach the OpenAI-compatible API to submit a legal image whose normalized pixels sum to zero when that image is…
When was the CVE record for CVE-2026-115914 published?
The CVE record for CVE-2026-115914 was published on October 26, 2026. That is the record's publication date; the vulnerability itself may have been made public earlier. EchelonGraph's copy of the record carries a last-modified date of October 30, 2026; the record at its source may have been updated since.
Is CVE-2026-115914 actively exploited?
CVE-2026-115914 is not currently on CISA's Known Exploited Vulnerabilities catalog. FIRST EPSS estimates a 0.4% probability of exploitation in the next 30 days (33rd percentile of EPSS-scored CVEs).
What is the CVSS score of CVE-2026-115914?
CVE-2026-115914 has a CVSS base score of 6.5 (a secondary CVSS source that NVD displays; NVD's own analysis pending).
How do I remediate CVE-2026-115914?
A fix for CVE-2026-115914 is available: update to the fixed version the vendor names in its advisory. The vendor advisories EchelonGraph has for CVE-2026-115914 are linked in the Vendor Advisories panel on this page.
Dependency Blast Radius
See which npm, PyPI, Go, and Maven packages are affected by CVE-2026-115914
Is Your Infrastructure Affected by CVE-2026-115914?
EchelonGraph automatically scans your cloud infrastructure and maps CVE exposure using blast radius analysis.