Project AURIONAI Use-Cases Research & Innovation
Tsinghua University
vision

Dinomaly2

Research model for industrial visual anomaly detection; image AUROC MVTec AD 99.9, VisA 99.3, Real-IAD 92.1. Licence unverified.

Figure from the Dinomaly2 paper: Overview: task settings, benchmarks, and scalability of Dinomaly2.
Levelmodel
Classvision
Parametersnot in the notes
Contextnot in the notes
Released24 Oct 2025
Opennessunknown
Runs onnot in the notes
Statusnot in the notes

Benchmark Results

vendor-reported
MVTec ADindustrial
Dinomaly2-L, multi-class unified, image AUROCvendor-reported
99.9
050100

Score in AUROC. Higher is better.

24 Oct 2025 arxiv.org
Real-IADindustrial
Dinomaly2-L, multi-class unified, image AUROCvendor-reported
92.1
050100

Score in AUROC. Higher is better.

24 Oct 2025 arxiv.org
VisAindustrial
Dinomaly2-L, multi-class unified, image AUROCvendor-reported
99.3
050100

Score in AUROC. Higher is better.

24 Oct 2025 arxiv.org
Where this comes from
Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).

Lineage

successor
Beijing Institute of TechnologyDinomaly is the parent (successor)medium confidencenotes
Where this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. The tree shows two generations each way, 2 models. Confidence and evidence are the notes' own.

Industry Use Cases

R&D Topics

Identity

Developer
Tsinghua University, Beijing Institute of Technology, Shanghai Jiao Tong University, Fudan University
Family
not in the notes
Released
24 Oct 2025 sourcesnotes
Weights published
–
Record checked
10 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture
dense transformer ViT-S/B/L on DINOv2-R features; multi-class unified unsupervised anomaly detection
Total parameters
– sources
Active parameters
–
Checkpoint parameters
–
Context length
not in the notes
Input
image
Output
not in the notes
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness
unknown
Licence
not in the notes sources
Access routes
research only
Where this comes from
The licence file governs over registry tags and aggregator labels.

Deployment

Runs on
not in the notes
Footprint
not in the notes
Runtimes
not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support
not stated
Regional languages
not in the notes
Jakarta hosting
unknown
Languages
not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.

Sources

  1. arxiv.org/pdf/2510.17611 https://arxiv.org/pdf/2510.17611

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Dinomaly2", https://aurion.semai.id/models/tsinghua-university-dinomaly2/, records of October 2026.