New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
Models Released in 2018
Browse every generative AI model released in 2018 — large language models, image generators, code models and more — ordered by release date. Compare their specs and rankings on our benchmarks page, or put any two side by side with compare. See also models from 2017 and models from 2019.
ResNeXt-101 32x48d
DNCON2
🇺🇸 University of Missouri
TF-LM-discourse LSTM (PTB)
🇺🇸 ESAT - PSI
TF-LM-discourse LSTM (WT2)
🇺🇸 ESAT - PSI
RNNLM + Dynamic KL Regularization (WT2)
🇺🇸 Northwestern University
RNMT+
🇺🇸 Google AI
Diffractive Deep Neural Network
🇺🇸 University of California Los Angeles (UCLA)
YOLOv3
🇺🇸 University of Washington
LSTM (Hebbian, Cache, MbPA)
🇬🇧 DeepMind
4 layer QRNN (h=2500)
🇺🇸 Salesforce Research
Chinese - English translation
🇺🇸 Microsoft
Residual Dense Network
🇺🇸 Northeastern University
Spectrally Normalized GAN
🇯🇵 Preferred Networks Inc
Multipop Adaptive Continuous Stack (PTB)
🇬🇧 DeepMind
TCN (13M)
🇺🇸 Carnegie Mellon University (CMU)
TCN (P-MNIST)
🇺🇸 Carnegie Mellon University (CMU)
ENAS
🇺🇸 Google Brain
DeepLabV3+
AmoebaNet-A (F=448)
🇺🇸 Google Brain
IMPALA
🇬🇧 DeepMind
ELMo
🇺🇸 University of Washington
QRNN
🇺🇸 Salesforce Research
T-DMCA
🇺🇸 Google Brain
DenseNet201
🇨🇳 Tsinghua University
ULM-FiT
🇺🇸 University of San Francisco
Refined Part Pooling
🇨🇳 Tsinghua University
RNNLM + Dynamic KL Regularization
🇺🇸 Northwestern University
2018 was another fast-moving year for generative AI. The models listed above span large language models (LLMs), image and video generators, code models and more, each with its own parameter count, context window, licensing and availability. To understand how the models of 2018 actually perform, compare their scores on our benchmarks page, and use compare to evaluate any two releases side by side.
Tracking AI by release year makes it easy to see how the frontier moves. Browse AI models from 2017 or AI models from 2019 to compare how capabilities, context windows and open-weights availability evolved year over year.