The AI Museum
AI Analytics1,072 recordsRecounted 23 September 2026

Language models in 2018

Weightsopenclosedunknownfrontier only
Year20262025202420232022202120202019201820172016
2 filters active, showing 24 of 1,072 recordsClear all

24

Models

19 organisations

24%

Open weights

5 open, 16 closed

0

Frontier

By Epoch's flag

5B

Largest traceable

Mesh-TensorFlow Transformer 4.9B (language)

2018

Span

10 with a traceable size

Releases

by month
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sept
Oct
Nov
Dec

Organisation

Country

Domain

Records

24
Language models in 2018, newest first. Each row links to the source Epoch AI records for it.
ModelOrganisationReleasedParametersWeights
Fine-tuned-AWD-LSTM-DOC (fin)Samsung R&D Institute Russia2018-1146MClosed
GPipe (Transformer)Google2018-116B *Not recorded
Mesh-TensorFlow Transformer 2.9B (translation)Google Brain2018-113B *Closed
Mesh-TensorFlow Transformer 4.9B (language)Google Brain2018-115BClosed
Multi-cell LSTMUniversity of Hyderabad2018-117M *Closed
BERT-LargeGoogle2018-10340MOpen
MemoReaderSamsung2018-10—Closed
TrellisNetCarnegie Mellon University (CMU)2018-10180MClosed
AWD-LSTM-MoS + dynamic evaluation (WT2, 2018)Peking University2018-0935MClosed
LSTM+NeuralCacheKU Leuven2018-092M *Closed
Transformer (Adaptive Input Embeddings) WT103Facebook AI Research2018-09247MOpen
Transformer + Simple Recurrent UnitASAPP2018-0990MClosed
(ensemble): AWD-LSTM-DOC (fin) × 5 (WT2)NTT Communication Science Laboratories2018-08185MOpen
AWD-LSTM-MoS+PDR + dynamic evaluation (WT2)IBM2018-0835MClosed
Big Transformer for Back-TranslationFacebook AI Research2018-08—Open
GPT-1OpenAI2018-06117M *Open
Relational Memory CoreDeepMind2018-06—Closed
aLSTM(depth-2)+RecurrentPolicy (WT2)University of Manchester2018-0532MClosed
Dropout-LSTM+Noise(Bernoulli) (WT2)Columbia University2018-0551M *Closed
4 layer QRNN (h=2500)Salesforce Research2018-03151M *Closed
Chinese - English translationMicrosoft2018-03—Not recorded
ELMoUniversity of Washington2018-0294M *Not recorded
QRNNSalesforce Research2018-02135M *Closed
T-DMCAGoogle Brain2018-01—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