The AI Museum
AI Analytics1,072 recordsRecounted 23 September 2026

Closed-weight vision models

Weightsopenclosedunknownfrontier only
Year20262025202420232022202120202019201820172016
2 filters active, showing 58 of 1,072 recordsClear all

58

Models

33 organisations

0%

Open weights

0 open, 58 closed

6

Frontier

By Epoch's flag

22B

Largest traceable

ViT-22B

1976–2025

Span

24 with a traceable size

Releases

by year, from 2010
2010201120122013201420152016201720182019202020212022202320242025

Organisation

Country

Domain

Records

newest 50 of 58
Closed-weight vision models, newest first. Each row links to the source Epoch AI records for it.
ModelOrganisationReleasedParametersWeights
Gemini Robotics-ER 1.5Google DeepMind2025-09—Closed
Seed1.5-VLByteDance2025-05—Closed
Computer-Using Agent (CUA)OpenAI2025-01—Closed
SeedEditByteDance2024-11—Closed
NVLM-H 72BNVIDIA2024-1072B *Closed
NVLM-X 72BNVIDIA2024-1072B *Closed
Harrison.rad.1Harrison.ai2024-09—Closed
BASIC-L + LionGoogle2023-023BClosed
ViT-22BGoogle2023-0222BClosed
CoCaGoogle Research2022-062BClosed
BASIC-LGoogle2021-113B *Closed
FlorenceMicrosoft2021-11893MClosed
ViT-G/14 (LiT)Google Research2021-113BClosed
CoAtNetGoogle2021-062BClosed
ViT-G/14Google Brain2021-062BClosed
Meta Pseudo LabelsGoogle Brain2021-03480M *Closed
Big Transfer (BiT-L)Google Brain2019-12928M *Closed
Noisy Student (L2)Carnegie Mellon University (CMU)2019-11480MClosed
AlphaX-1Facebook AI Research2019-105MClosed
CPC v2DeepMind2019-05303M *Closed
Neuro-Symbolic Concept LearnerMassachusetts Institute of Technology (MIT)2019-04—Closed
YOLOv3University of Washington2018-0457M *Closed
DL scaling ImageBaidu2017-12121MClosed
VQ-VAEDeepMind2017-11—Closed
SRGANTwitter2017-05—Closed
GAN-AdvancerOpenAI2016-12—Closed
DAC-CSRJiangnan University2016-11—Closed
DTN (Domain Transfer Network)Facebook AI Research2016-11—Closed
GAWWNUniversity of Michigan2016-10—Closed
ResNet-200Microsoft Research Asia2016-09—Closed
CCLSenseTime2016-06—Closed
CMS-RCNNIEEE2016-06138M *Closed
Template AdaptationUniversity of Oxford2016-03138M *Closed
3DDFAChinese Academy of Sciences2015-115MClosed
SAF R-CNNBeijing Institute of Technology2015-10138M *Closed
DCNNUniversity of Maryland2015-085MClosed
CompACT-DeepUniversity of California San Diego2015-07—Closed
Deep CNN + COTSIEEE2015-075MClosed
VGG-FaceUniversity of Oxford2015-01138MClosed
ADAM (CIFAR-10)University of Amsterdam2014-122MClosed
TA-CNNChinese University of Hong Kong (CUHK)2014-11706,048 *Closed
ACF-WIDERChinese Academy of Sciences2014-076,144Closed
Dropout: SVHNUniversity of Toronto2014-0648MClosed
Hierarchical Scene Labeling (Stanford Background)New York University (NYU)2013-0852MClosed
DNN EM segmentationIDSIA2012-12218,896Closed
Dropout (CIFAR)University of Toronto2012-06—Closed
Dropout (ImageNet)University of Toronto2012-06—Closed
Dropout (MNIST)University of Toronto2012-066MClosed
CNN committee (traffic sign)IDSIA2011-101M *Closed
CNN Committee (MNIST)IDSIA2011-09120,620 *Closed

An asterisk marks a parameter figure the source does not consider traceable to the people who built the model. That judgement is the source’s and it has false negatives: Kimi K3 is marked speculative at 2.8T while Moonshot’s own model card states the figure. Why that distinction carries the whole exhibit.

Counts are of models Epoch AI records as notable, not of every model released. The most recent month is always incomplete, because a model enters the dataset when it is reviewed rather than when it ships.

Epoch AI, 'Data on AI Models'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/data/ai-models-documentation' Used under CC BY 4.0.

How this museum is made