chapter2-training-compute.csv

# Training compute of notable AI models
# Derived from: data/datasets/sources/epoch-notable-models.csv (retained raw copy)
# Upstream: Epoch AI, "Data on AI Models", https://epoch.ai/data/ai-models-documentation
#   Daily-updated dataset, CC BY.
# Citation key: EpochAI2025AIModels
# Verified 2026-09-03: all 529 rows trace to a raw record by model name
#   (raw holds 1,036 models; this receipt is the subset carrying a training
#   compute figure). The "frontier" column is Epoch's own "Frontier model" field
#   (top 10 by training compute when released), not a running maximum.
# Re-verified 2026-09-29: 529/529 rows match the raw copy; against the live Epoch
#   file, 523 rows are identical and 6 have since been revised by Epoch (4 dates,
#   2 compute values withdrawn). The chapter's fit (2010 to 2026.6, 467 rows) gives
#   10**slope = 4.35 per year; it is the book's own fit, not an Epoch figure.
model,publication_date,training_compute_flop,frontier
Inkling,2026-07-15,1.845e+24,0
Nemotron 3 Ultra,2026-06-04,6.6e+24,0
Composer 2.5,2026-05-18,3.87e+25,0
DeepSeek-V4-Pro,2026-04-24,9.702e+24,0
DeepSeek-V4-Flash,2026-04-24,2.496e+24,0
MiMo-V2.5-Pro,2026-04-23,6.804e+24,0
EXAONE 4.5,2026-04-09,3.85e+24,0
Composer 2,2026-03-19,2.32e+25,0
GLM-5,2026-02-17,6.84e+24,0
Kimi K2.5,2026-02-02,5.8e+24,0
K-EXAONE,2025-12-31,1.52e+24,0
GLM-4.7,2025-12-22,4.42e+24,0
Nemotron 3-Nano-30B-A3B,2025-12-15,4.8e+23,0
Olmo 3,2025-11-20,1.1e+24,0
Kimi K2 Thinking,2025-11-06,4.2e+24,0
Ling-1T,2025-10-10,6.000001e+24,0
GLM-4.6,2025-09-30,4.42e+24,0
Qwen3-Omni-30B-A3B,2025-09-22,3.6e+22,0
AgentFounder-30B,2025-09-16,6.5367e+23,0
Qwen3-Max,2025-09-05,1.512e+25,0
LongCat-Flash,2025-09-01,3.726e+24,0
GPT-5,2025-08-07,6.6e+25,0
gpt-oss-120b,2025-08-05,4.94e+24,0
gpt-oss-20b,2025-08-05,5.49e+23,0
GLM-4.5,2025-08-05,4.42e+24,0
Qwen3-235B-A22B-Thinking (Jul 2025),2025-07-25,4.752e+24,0
Qwen3-235B-A22B (Jul 2025),2025-07-25,4.752e+24,0
Qwen3-Coder-480B-A35B,2025-07-22,1.575e+24,0
EXAONE 4.0 (32B),2025-07-15,2.69000000000001e+24,0
Kimi K2,2025-07-11,2.976e+24,0
Grok 4,2025-07-09,5.0000000000001e+26,1
FGN,2025-06-12,9.618950880000001e+21,0
DeepSeek-R1 (May 2025),2025-05-28,4.020010000000001e+24,0
Seed1.5-VL,2025-05-11,1.388556e+24,0
Qwen3-235B-A22B,2025-04-29,4.752e+24,0
Pangu Ultra,2025-04-10,1.0692e+25,0
Llama 4 Scout,2025-04-05,4.08e+24,0
Llama 4 Maverick,2025-04-05,2.244000000001e+24,0
Llama 4 Behemoth (preview),2025-04-05,5.18400000000001e+25,1
DeepSeek-V3 (Mar 2025),2025-03-24,3.3e+24,0
EXAONE Deep 32B,2025-03-16,1.26e+24,0
Hunyuan-TurboS,2025-03-11,5.4e+24,0
QwQ-32B,2025-03-06,3.51e+24,0
GPT-4.5,2025-02-27,3.8e+26,1
Claude 3.7 Sonnet,2025-02-24,3.35e+25,0
Grok 3,2025-02-17,3.5e+26,1
Eagle 2,2025-01-20,4.7156e+22,0
DeepSeek-R1,2025-01-20,3.5e+24,0
DeepSeek-V3,2024-12-24,3.3e+24,0
EXAONE 3.5 32B,2024-12-09,1.25e+24,0
Llama 3.3 70B,2024-12-06,6.8649768e+24,0
Amazon Nova Pro,2024-12-03,6.000010000000001e+24,0
Hunyuan-Large,2024-11-06,3.49237e+24,0
Doubao-pro,2024-10-28,2.505e+25,0
NVLM-X 72B,2024-10-22,3.0398181e+24,0
NVLM-H 72B,2024-10-22,3.02e+24,0
NVLM-D 72B,2024-10-22,3.02e+24,0
Yi-Lightning,2024-10-18,1.5e+24,0
CHAI-1,2024-10-15,7.7605724e+21,0
RDT-1B,2024-10-10,4.06e+22,0
Movie Gen Video,2024-10-04,1.65e+24,0
Llama 3.2 11B,2024-09-24,5.79e+23,0
Telechat2-115B,2024-09-20,6.9e+24,0
Qwen2.5-72B,2024-09-19,7.8e+24,0
Qwen2.5 Instruct (72B),2024-09-19,7.8516e+24,0
Qwen2.5-32B,2024-09-17,3.51e+24,0
DeepSeek-V2.5,2024-09-06,1.7892e+24,0
Grok-2,2024-08-13,2.96e+25,1
LLaVA-OV-72B,2024-08-06,3.036551985824e+24,0
AFM-server,2024-07-29,4.3e+24,0
AFM-on-device,2024-07-29,4.5126e+23,0
Mistral Large 2,2024-07-24,2.13e+25,0
Llama 3.1-405B,2024-07-23,3.8e+25,1
ESM3 (98B),2024-06-25,1.07e+24,0
Claude 3.5 Sonnet,2024-06-20,2.700000000000001e+25,1
DeepSeek-Coder-V2 236B,2024-06-17,1.2852e+24,0
Nemotron-4 340B,2024-06-14,1.8e+25,1
OpenVLA,2024-06-13,1.1e+23,0
Llama-3.1-Nemotron-70B-Instruct,2024-06-12,7.929e+24,0
Qwen2-72B,2024-06-07,3.02e+24,0
ALLaM adapted 70B,2024-05-21,1.062e+24,0
Octo-Base,2024-05-20,5.85e+20,0
Yi-Large,2024-05-13,1.8e+24,0
AlphaFold 3,2024-05-08,4.1405645e+22,0
VILA1.5-13B,2024-05-03,2.3003136e+21,0
GenCast,2024-05-01,8.169984e+20,0
Llama 3-70B,2024-04-18,7.861e+24,0
Reka Core,2024-04-15,8.400010000000001e+24,0
DBRX,2024-03-27,2.6e+24,0
MM1-30B,2024-03-14,4.86e+23,0
Inflection-2.5,2024-03-07,8.000001e+24,0
Aramco Metabrain AI,2024-03-04,1.05e+25,0
Mistral Large,2024-02-26,1.12e+25,0
MegaScale (Production),2024-02-23,3.9e+24,0
Stable Diffusion 3,2024-02-22,5.0000000000000004e+22,0
Qwen1.5-72B,2024-02-04,1.3e+24,0
GQA-8-XXL,2023-12-23,3.4912896e+22,0
nekomata-14b,2023-12-21,2.5562e+23,0
FunSearch,2023-12-14,3.87e+23,0
CogAgent,2023-12-14,6.707e+22,0
VILA-13B,2023-12-12,2.3003136e+21,0
Mixtral 8x7B,2023-12-11,7.74e+23,0
Llama Guard,2023-12-07,1.6e+23,0
Gemini 1.0 Ultra,2023-12-06,5.0000000001e+25,1
Qwen-72B,2023-11-30,1.3e+24,0
Inflection-2,2023-11-22,1.001e+25,1
Nemotron-3-8B,2023-11-15,1.8e+23,0
GraphCast,2023-11-14,2.1000000000000002e+22,0
Volcano 13B,2023-11-13,4.56e+22,0
SPHINX (Llama 2 13B),2023-11-13,3.04e+22,0
RoFormer,2023-11-08,2.162688e+18,0
MultiBand Diffusion,2023-11-08,2.6e+19,0
CogVLM-17B,2023-11-06,6.331e+22,0
LLaVA 1.5,2023-11-05,7.807e+22,0
Grok-1,2023-11-04,2.90000000001e+24,0
Yi-34B,2023-11-02,6.1e+23,0
Skywork-13B,2023-10-30,2.5e+23,0
ChatGLM3-6B,2023-10-27,5.04e+22,0
CODEFUSION (Python),2023-10-26,7.92e+18,0
RoseTTAFold All-Atom (RFAA),2023-10-09,2.14e+20,0
FinGPT-13B,2023-10-07,1.6e+23,0
Amazon Titan,2023-09-28,4.8e+24,1
Falcon-180B,2023-09-06,3.76e+24,1
Swift,2023-08-30,5.337e+16,0
Jais,2023-08-29,4.8946763e+22,0
PeptideBERT,2023-08-28,4.9e+16,0
GGNN,2023-08-05,7.56e+21,0
AudioLM,2023-07-26,3.9e+18,0
Llama 2-70B,2023-07-18,8.1e+23,0
Llama 2-7B,2023-07-18,8.4e+22,0
Claude 2,2023-07-11,3.866e+24,1
xTrimoPGLM -100B,2023-07-06,6.2e+23,0
InternLM,2023-07-06,9.984e+23,0
Pangu-Weather,2023-07-05,3.98e+22,0
HyenaDNA,2023-06-27,1.811e+21,0
GPT-4 (Jun 2023),2023-06-13,2.1e+25,1
PaLI-X,2023-05-29,5.6e+23,0
ONE-PEACE,2023-05-18,1.8e+20,0
InstructBLIP,2023-05-11,1.94e+20,0
PaLM 2,2023-05-10,7.34e+24,1
StarCoder,2023-05-09,8.46e+22,0
LLaVA,2023-04-17,7.8049e+22,0
DINOv2,2023-04-14,7.41851136e+21,0
Incoder-6.7B,2023-04-09,3.00001e+21,0
Segment Anything Model,2023-04-05,7.8e+21,0
BloombergGPT,2023-03-30,2.36e+23,0
VideoMAE V2,2023-03-29,9.7e+21,0
SigLIP 400M,2023-03-27,4.9467301e+21,0
PanGu-Σ,2023-03-20,4.67e+23,0
GPT-4 (Mar 2023),2023-03-15,2.1e+25,1
Falcon-40B,2023-03-15,2.4e+23,0
AudioGen,2023-03-05,9.5e+21,0
DiT-XL/2,2023-03-02,6e+20,0
LLaMA-65B,2023-02-24,5.5e+23,0
ViT-22B,2023-02-10,1.93248e+23,0
BLIP-2 (Q-Former),2023-01-30,1.20000000001e+21,0
DDPM-IP (CelebA),2023-01-27,3.5e+20,0
Ankh_large,2023-01-16,6.5e+21,0
Nucleotide Transformer,2023-01-15,8.08e+21,0
DreamerV3,2023-01-10,2.2032e+20,0
VALL-E,2023-01-05,1.01e+19,0
Hybrid H3-2.7B,2022-12-28,6.48e+21,0
CaLM,2022-12-19,2.9e+19,0
Vega v2,2022-12-04,7.76e+22,0
Discriminator Guidance,2022-11-28,2.1570000001e+20,0
AR-LDM,2022-11-20,5.1e+20,0
Fusion in Encoder,2022-11-18,1.3e+20,0
Galactica,2022-11-16,3.24e+23,0
EVA-01,2022-11-14,1.501e+22,0
InternImage,2022-11-10,2.408e+21,0
Mogrifier RLSTM (WT2),2022-11-03,1.4e+17,0
eDiff-I,2022-11-02,5.46e+19,0
U-PaLM (540B),2022-10-20,2.53e+24,1
Flan-PaLM 540B,2022-10-20,2.540000000001e+24,1
GenSLM,2022-10-11,1.42e+21,0
AlphaTensor,2022-10-05,7.1414784e+20,0
DiffDock,2022-10-04,7.2e+19,0
Whisper,2022-09-21,4.2072663e+21,0
PaLI,2022-09-14,1.69e+23,0
BEIT-3,2022-08-22,7e+19,0
BlenderBot 3,2022-08-10,4.3e+23,0
GLM-130B,2022-08-04,3.5490054945e+23,0
AlexaTM 20B,2022-08-02,2.04374016e+23,0
OmegaPLM,2022-07-22,1.03514112e+22,0
ESM2-15B,2022-07-21,7.35000000001e+22,0
BLOOM-176B,2022-07-11,3.65664e+23,0
NLLB,2022-07-06,1.751113728e+22,0
CodeT5-large,2022-07-05,2.72e+21,0
Minerva (540B),2022-06-29,2.7415e+24,1
ProGen2-xlarge,2022-06-27,1.35e+22,0
Parti,2022-06-22,5.09607936e+23,0
CoCa,2022-06-14,7.3e+22,0
DITTO,2022-06-06,3.31776e+18,0
Tranception,2022-05-27,7.24e+21,0
GPT-2 Medium (FlashAttention),2022-05-27,8.9280922e+20,0
Imagen,2022-05-23,1.46e+22,0
Gato,2022-05-12,4.02e+21,0
UL2,2022-05-10,1.2e+23,0
OPT-175B,2022-05-02,4.3e+23,0
Flamingo,2022-04-29,2.18972000000001e+23,0
Sparse all-MLP,2022-04-14,5.32224e+20,0
Stable Diffusion (LDM-KL-8-G),2022-04-13,5.0000000000000004e+22,0
BERT-RBP,2022-04-07,1.4e+20,0
DALL·E 2,2022-04-06,3.3695784e+23,0
PaLM (540B),2022-04-04,2.5272e+24,1
Chinchilla,2022-03-29,5.76e+23,0
Make-A-Scene,2022-03-24,6.4172851e+21,0
"Segatron-XL large, M=384 + HCP",2022-03-21,2.65e+19,0
GPT-3.5 (davinci-002),2022-03-15,2.578e+24,1
ViT-G (model soup),2022-03-10,3.4e+21,0
Statement Curriculum Learning,2022-03-02,1.7901648e+22,0
PolyCoder,2022-02-26,1.1e+21,0
FourCastNet,2022-02-22,3.4504704e+20,0
ST-MoE,2022-02-17,2.9e+23,0
ProteinBERT,2022-02-10,6.5e+19,0
LaMDA,2022-02-10,3.55e+23,0
GPT-NeoX-20B,2022-02-09,9.31627008e+22,0
RETRO-7B,2022-02-07,1.68e+22,0
AlphaCode,2022-02-02,2.38010000000001e+23,0
InstructGPT 175B,2022-01-27,3.19181e+23,0
Detic,2022-01-07,2.34399744e+19,0
ERNIE 3.0 Titan,2021-12-23,1.0421e+24,1
XGLM-7.5B,2021-12-20,2.25e+22,0
Contriever,2021-12-16,1.57e+20,0
GLaM,2021-12-13,3.6363112434e+23,0
Gopher (280B),2021-12-08,6.31e+23,1
Student of Games,2021-12-06,3.6679273004682866e+22,0
NÜWA,2021-11-24,7.24598784e+21,0
Florence,2021-11-22,4.831e+22,0
BASIC-L,2021-11-19,4.12e+22,0
Swin Transformer V2 (SwinV2-G),2021-11-18,1.1e+21,0
Masked Autoencoders ViT-H,2021-11-11,4.6e+20,0
Projected GAN,2021-11-01,1.05e+19,0
CodeT5-base,2021-11-01,1.56e+21,0
S4,2021-10-31,7.8328627e+19,0
base LM+GNN+kNN,2021-10-17,5.2587456e+19,0
Yuan 1.0,2021-10-12,3.5380000000001e+23,1
Megatron-Turing NLG 530B,2021-10-11,8.586e+23,1
AlphaFold-Multimer,2021-10-04,4.35e+21,0
Turing ULRv5,2021-09-28,2.8983951e+22,0
PLATO-XL,2021-09-20,9.9e+21,0
HyperCLOVA 204B,2021-09-10,2.0000000001e+23,1
PermuteFormer,2021-09-06,2.775e+18,0
FLAN 137B,2021-09-03,2.047e+24,1
XLMR-XXL,2021-08-17,3.366e+22,0
DNABERT,2021-08-15,1.07e+20,0
Zidong Taichu,2021-08-11,8.016e+20,0
Jurassic-1-Jumbo,2021-08-11,3.7e+23,1
YOLOX-X,2021-08-06,6.34275e+20,0
SEER,2021-07-29,1.8e+22,0
HuBERT,2021-07-27,5.54e+21,0
GOAT,2021-07-27,2.412e+22,0
Codex,2021-07-07,7.344e+22,0
ERNIE 3.0,2021-07-05,2.25e+22,0
Adaptive Input Transformer + RD,2021-06-28,8.58e+19,0
Fold2Seq,2021-06-24,1.4e+17,0
EfficientNetV2-XL,2021-06-23,9.56e+19,0
StyleGAN3-T,2021-06-21,1.70208e+21,0
StyleGAN3-R,2021-06-21,2.42784e+21,0
Denoising Diffusion Probabilistic Models (LSUN Bedroom),2021-06-11,7.840125000000001e+19,0
ALIGN,2021-06-11,2.598670000001e+22,0
DeBERTa,2021-06-10,2.588e+22,0
EMDR,2021-06-09,1.91e+21,0
CoAtNet,2021-06-09,4.27e+22,0
ViT-G/14,2021-06-08,5.85e+22,0
ByT5-XXL,2021-05-28,8.1e+22,0
Transformer local-attention (NesT-B),2021-05-26,2.40576e+19,0
CogView,2021-05-26,2.68e+22,0
MedBERT,2021-05-20,9.47e+18,0
ADM,2021-05-11,6.2e+21,0
ProtT5-XL-U50,2021-05-04,1.8704498688e+22,0
ProtBERT-BFD,2021-05-04,3.9e+22,0
PLUG,2021-04-19,3.5997696e+22,0
M6-T,2021-03-05,5.5e+21,0
Generative BST,2021-03-05,1.449e+22,0
Meta Pseudo Labels,2021-03-01,4.79e+22,0
SRU++ Large,2021-02-24,2.1173704e+19,0
MSA Transformer,2021-02-13,5.49e+21,0
DLWP,2021-02-09,5.6845152e+18,0
DeiT-B,2021-01-15,7.884e+19,0
Switch,2021-01-11,8.22e+22,1
DALL-E,2021-01-05,4.7e+22,0
CLIP (ViT L/14@336px),2021-01-05,1.05e+22,0
ERNIE-Doc (247M),2020-12-31,3.0302798e+19,0
CT-MoS (WT2),2020-12-25,5.4e+17,0
DensePhrases,2020-12-23,2.09952e+18,0
ESM1b,2020-12-15,5.1e+21,0
CPM-Large,2020-12-01,2.6052e+20,0
AlphaFold 2,2020-11-30,2.99e+21,0
KEPLER,2020-11-23,1.66e+21,0
wave2vec 2.0 LARGE,2020-10-22,3.87072e+21,0
ViT-Huge/14,2020-10-22,4.262e+21,0
German ELECTRA Large,2020-10-21,1.42829568e+21,0
mT5-XXL,2020-10-20,8.2e+22,1
Conformer + Wav2vec 2.0 + Noisy Student,2020-10-20,7.6e+21,0
LUKE,2020-10-02,1.8144e+22,0
ProBERTa,2020-09-01,9.72e+18,0
ERNIE-GEN (large),2020-08-06,2e+20,0
DeLighT,2020-08-03,3.8016e+18,0
GShard (dense),2020-06-30,4.765e+22,1
GPT-3 175B (davinci),2020-05-28,3.14e+23,1
DETR,2020-05-26,4e+20,0
UnifiedQA,2020-05-02,1.65e+19,0
Once for All,2020-04-29,6.237e+20,0
MetNet,2020-03-24,9.510912e+18,0
ELECTRA,2020-03-23,3.1e+21,0
Tensor-Transformer(1core)+PN (WT103),2020-03-17,1.58e+18,0
TransformerXL + spectrum control,2020-03-11,2.6289761e+19,0
Feedback Transformer,2020-02-21,7.690547e+18,0
FFN SwiGLU,2020-02-14,3.3687317e+19,0
Turing-NLG,2020-02-13,1.57e+22,0
ALBERT-xxlarge,2020-02-09,2.39e+21,0
TaLK Convolution,2020-02-08,2.6990346e+19,0
Meena,2020-01-28,1.12e+23,1
ContextNet + Noisy Student,2020-01-19,8.16e+21,0
AlphaFold,2020-01-15,1e+20,0
DD-PPO,2019-12-19,7.8e+20,0
OpenAI Five Rerun,2019-12-13,1.3e+22,0
OpenAI Five,2019-12-13,6.7e+22,1
MMLSTM (WT-2),2019-12-05,1.938e+17,0
MMLSTM (PTB),2019-12-05,5.8298782e+16,0
Transformer-XL DeFINE (141M),2019-11-27,1.74276e+18,0
MuZero,2019-11-19,4.8e+19,0
Noisy Student (L2),2019-11-11,2.612e+22,1
Sandwich Transformer,2019-11-10,2.3504093e+19,0
CamemBERT,2019-11-10,8.3e+20,0
XLM-RoBERTa,2019-11-05,2.076e+22,1
Base LM + kNN LM + Continuous Cache,2019-11-01,3.05292e+19,0
AlphaStar,2019-10-30,1.0773400001e+23,1
T5-11B,2019-10-23,3.3e+22,1
T5-3B,2019-10-23,9.0000000001e+21,1
DistilBERT,2019-10-02,1.24416e+19,0
AlphaX-1,2019-10-02,8.89344e+17,0
Megatron-LM (8.3B),2019-09-17,9.1e+21,1
Megatron-LM (1.2B),2019-09-17,1.1304e+21,0
Megatron-BERT,2019-09-17,2.2e+22,1
UDSMProt,2019-09-04,6.37e+17,0
trRosetta,2019-08-22,3.8047968e+19,0
Pluribus,2019-07-11,6.6e+16,0
RoBERTa Large,2019-07-01,8.5067e+21,1
AWD-LSTM + MoS + Partial Shuffled,2019-06-10,3.15e+17,0
Transformer-XL Large + Phrase Induction,2019-06-04,3.7848651e+20,0
XLNet,2019-06-01,6.19e+21,1
DLRM-2020,2019-05-31,4e+18,0
AWD-LSTM-DRILL + dynamic evaluation† (WT2),2019-05-14,4.08e+17,0
MuseNet,2019-04-25,2.208301056e+20,0
BERT-Large-CAS (PTB+WT2+WT103),2019-04-20,1.5405e+20,0
WeNet (Penn Treebank),2019-04-08,7.30000001e+17,0
Cross-lingual alignment,2019-04-04,2.56e+18,0
SciBERT,2019-03-26,8.926848e+19,0
KataGo,2019-02-27,2.32e+19,0
GPT-2 (1.5B),2019-02-14,1.920000000001e+21,0
Hanabi 4 player,2019-02-01,4.3e+18,0
Transformer-XL (257M),2019-01-09,3.7832771e+20,0
StyleGAN,2018-12-12,3.93e+16,0
Multi-cell LSTM,2018-11-15,2006640000000000.0,0
Fine-tuned-AWD-LSTM-DOC (fin),2018-11-12,5.188e+16,0
Mesh-TensorFlow Transformer 4.9B (language),2018-11-05,1.617408e+20,0
Mesh-TensorFlow Transformer 2.9B (translation),2018-11-05,6.84288e+19,0
TrellisNet,2018-10-15,2.78e+18,0
BERT-Large,2018-10-11,2.85e+20,0
Transformer (Adaptive Input Embeddings) WT103,2018-09-28,4.47e+19,0
LSTM+NeuralCache,2018-09-24,982800000000000.0,0
Transformer + Simple Recurrent Unit,2018-09-17,1.1e+19,0
(ensemble): AWD-LSTM-DOC (fin) × 5 (WT2),2018-08-30,6.66e+17,0
Big Transformer for Back-Translation,2018-08-28,4.7808e+20,0
Big-Little Net (speech),2018-07-10,4.290048e+17,0
Big-Little Net,2018-07-10,2.46048e+17,0
FTW (For The Win),2018-07-03,3.49e+19,0
GPT-1,2018-06-01,1.7578125e+19,0
aLSTM(depth-2)+RecurrentPolicy (WT2),2018-05-22,7.296e+16,0
Dropout-LSTM+Noise(Bernoulli) (WT2),2018-05-03,1.27e+17,0
ResNeXt-101 32x48d,2018-05-02,8.74395e+21,1
YOLOv3,2018-04-08,1.3416380824e+19,0
4 layer QRNN (h=2500),2018-03-22,5.9158815e+17,0
IMPALA,2018-02-05,1.68e+20,0
QRNN,2018-02-01,6.8866472e+17,0
ELMo,2018-02-01,3300100000000010.0,0
AlphaZero,2017-12-05,1.0600000001e+20,0
"AWD-LSTM-MoS + dynamic evaluation (WT2, 2017)",2017-11-10,3.36e+18,0
Fraternal dropout + AWD-LSTM 3-layer (WT2),2017-10-31,3.06e+17,0
AlphaGo Master,2017-10-19,3.4100000001e+20,0
AlphaGo Zero,2017-10-18,6.49439910290227e+20,1
AWD-LSTM+WT+Cache+IOG (WT2),2017-09-26,3180000000000000.0,0
ISS,2017-09-15,3400000000000000.0,0
PyramidNet,2017-09-06,2340000000000000.0,0
GL-LWGC-AWD-MoS-LSTM + dynamic evaluation (WT2),2017-08-29,4.56e+17,0
Libratus,2017-08-19,5.51e+20,0
EI-REHN-1000D,2017-08-14,1.06e+16,0
OpenAI TI7 DOTA 1v1,2017-08-11,6.046095222592002e+20,1
RetinaNet-R101,2017-08-07,2.065392e+18,0
AWD-LSTM - 3-layer LSTM (tied) + continuous cache pointer (WT2),2017-08-07,2.97e+17,0
ConvS2S (ensemble of 8 models),2017-07-25,5.64e+19,0
JFT,2017-07-10,8.43e+20,1
DeepLoc,2017-07-07,5.781024e+17,0
Transformer,2017-06-12,7.4245248e+18,0
MoE-Multi,2017-01-23,9.393905664e+19,0
DeepStack,2017-01-06,1.446336e+19,0
EnhanceNet,2016-12-23,1.3079231999999998e+17,0
HR-ResNet101,2016-12-13,7.077e+18,0
PolyNet,2016-11-17,6.4e+19,0
ResNeXt-101 (64×4d),2016-11-16,1.2e+19,0
NASv3 (CIFAR-10),2016-11-05,2.2e+21,1
NAS with base 8 and shared embeddings,2016-11-05,1.05e+16,0
BIDAF,2016-11-05,3.4686144e+18,0
VD-LSTM+REAL Large,2016-11-04,2.13e+16,0
SPIDER2,2016-10-28,1.822e+16,0
Xception,2016-10-07,4.36e+20,1
Pointer Sentinel-LSTM (medium),2016-09-26,7490000000000000.0,0
GNMT,2016-09-26,6.620000000001e+21,1
ResNet-200,2016-09-17,2.9741645e+19,0
R-FCN,2016-06-21,7.1935776e+17,0
Named Entity Recognition model,2016-03-04,9.69408e+16,0
AlphaGo Lee,2016-01-27,1.9e+21,1
"Variational (untied weights, MC) LSTM (Large)",2015-12-16,5886144000000000.0,0
ResNet-152 (ImageNet),2015-12-10,1.041408e+19,0
ResNet-101 (ImageNet),2015-12-10,7.004e+18,0
Inception v3,2015-12-02,1e+20,1
SAF R-CNN,2015-10-28,1.2311081250000001e+19,1
AlphaGo Fan,2015-10-01,3.8e+20,1
DCNN,2015-08-07,4.8098691e+17,0
U-Net,2015-05-18,5.0832252e+16,0
TC-DNN-BLSTM-DNN,2015-04-06,1.9410191999999997e+17,0
genCNN + dyn eval,2015-03-17,3.4153451e+16,0
"MSRA (C, PReLU)",2015-02-06,2.397403008e+19,1
ADAM (CIFAR-10),2014-12-22,624979800000000.0,0
Fractional Max-Pooling,2014-12-18,1e+17,0
SNM-skip,2014-12-03,2.97600000001e+20,1
TA-CNN,2014-11-29,1.0854e+16,0
GoogLeNet / InceptionV1,2014-09-17,1.51e+18,0
SPN-4+KN5,2014-09-14,4.4e+16,0
Seq2Seq LSTM,2014-09-10,5.6e+19,1
VGG19,2014-09-04,1.1e+19,1
VGG16,2014-09-04,1.2291e+19,1
RNNsearch-50*,2014-09-01,1.5552e+18,1
ACF-WIDER,2014-07-15,76380000000000.0,0
SmooCT,2014-07-01,6.9e+16,0
SPPNet,2014-06-18,3.411072e+18,1
GANs,2014-06-10,5.184e+17,0
Image generation,2013-12-20,475200000000000.0,0
DQN,2013-12-19,2846883840000000.0,0
TransE,2013-12-05,1.340928e+18,1
Visualizing CNNs,2013-11-12,5.32e+17,1
Word2Vec (large),2013-10-16,3.888e+16,0
RNTN,2013-10-01,1.422e+16,0
RCTM,2013-10-01,9331200000000000.0,0
Hierarchical Scene Labeling (Stanford Background),2013-08-01,2.3774688e+17,0
ReLU-Speech,2013-05-26,1.2773376e+17,0
DistBelief NNLM,2013-01-16,2.612736e+18,1
DNN EM segmentation,2012-12-03,4.78e+17,1
DistBelief Speech,2012-12-03,3.114e+17,0
AlexNet,2012-09-30,4.7e+17,1
LSTM LM,2012-09-09,1.66e+16,0
Unsupervised High-level Feature Learner,2012-07-12,6e+17,1
Dropout (MNIST),2012-06-03,6039370800000000.0,0
Dropout (ImageNet),2012-06-03,2.731968e+17,1
Dropout (CIFAR),2012-06-03,4268700000000000.0,0
CNN committee (traffic sign),2011-10-03,991981425600000.0,0
CNN Committee (NIST),2011-09-18,2.6e+16,0
CNN Committee (MNIST),2011-09-18,5.2e+16,1
High Performance CNN (NORB),2011-07-16,2.57985e+16,0
Deep Autoencoders,2011-04-29,3.672864e+16,1
RNN LM,2010-09-26,5.396e+16,1
Pooling CNN (NORB),2010-09-15,1456277227200000.0,0
Pooling CNN (Caltech 101),2010-09-15,1221124128768000.0,0
iCCCP,2010-06-13,1080000000000000.0,0
Feedforward NN,2010-05-13,350000000000000.0,0
LCNP NORB,2009-11-22,2452090060800000.0,0
LCNP MNIST,2009-11-22,4181852160000000.0,1
LCNP LabelMe,2009-11-22,3295150080000000.0,1
Two Stage Feature Extraction (MNIST),2009-09-01,20754000000000.0,0
GPU DBNs,2009-06-15,1000000000000000.0,0
GNN,2008-12-09,1614600000.0,0
SB-LM,2007-06-22,1.4494464e+18,1
KN-LM,2007-06-22,7.7303808e+17,1
SVM-CNN,2006-06-17,745200000000000.0,1
RankNet,2005-08-07,3482081588304.0,0
Hierarchical LM,2005-01-06,115848000000000.0,1
LMICA,2004-12-01,2782080000000000.0,1
Invariant CNN,2004-06-27,974230000000.0,0
NPLM (Brown),2003-03-15,132076260000000.0,1
NPLM (AP News),2003-03-15,1666869200000000.0,1
Decision tree (classification),2001-12-08,63000000000000.0,1
PoE MNIST,2000-11-28,51810000000000.0,1
Neural LM,2000-11-28,6339000000000000.0,1
RECONTRA-uncategorized,1999-06-02,3864000000000.0,0
RECONTRA-categorized,1999-06-02,8013600000000.0,1
LeNet-5,1998-11-01,2810937600000.0,1
LSTM,1997-11-15,31512000000000.0,1
SOM-CNN,1997-01-31,31431600000.0,0
System 11,1996-06-18,25859616000.0,0
MUSIC perceptron,1996-06-03,881733600000.0,1
LISSOM,1995-11-27,195544800000.0,0
Predictive Coding NN,1994-12-02,18621900000000.0,1
NeuroChess,1994-12-02,858730812676.0,1
Mixture of linear models,1994-12-02,453600000000.0,0
JPMAX,1994-12-02,80828280.0,0
Ceramic-MLP,1994-01-07,4531200000.0,0
ANN Eye Tracker,1993-11-29,17534400000.0,0
Siamese-TDNN,1993-08-01,12869570138112.0,1
Cancer drug mechanism prediction,1992-10-16,53460000.0,0
TD-Gammon,1992-05-01,18232157622832.703,1
Weight Decay,1991-12-02,75474000000.0,1
SexNet compression,1990-10-01,78775200000.0,1
Bankruptcy-NN,1990-06-17,3059337600.0,0
NETtalk reimplementation,1990-06-01,35811936000.0,1
Zip CNN,1989-12-01,1496338054440.0,1
Speaker-independent vowel classification,1989-11-27,7485696000.0,0
Handwritten digit recognition network,1989-11-27,181440000000.0,1
Invariant image recognition,1989-06-18,27000000000.0,1
MLN-ASR,1988-08-01,296425000.0,0
Translation-invariant MLP,1987-06-15,18032947200.0,1
NetTalk (transcription),1987-06-06,28328002560.0,1
NetTalk (dictionary),1987-06-06,27664065000.0,1
MLP with back-propagation,1986-10-01,673920000.0,1
Distributed representation NN,1986-08-15,388800000.0,1
ASE+ACE,1983-09-01,324000000.0,1
Neocognitron,1980-04-01,273738240.0,1
Cognitron,1975-09-01,5184000.0,0
LTE speaker verification system,1966-11-01,105917060.0,1
Heuristic Reinforcement Learning,1965-10-01,1080000.0,0
Print Recognition Logic,1963-01-01,22500000.0,1
Linear Decision Functions,1962-06-01,1559250.0,1
ADALINE,1960-06-30,6600.0,1
Perceptron (1960),1960-03-30,720000000.0,1
Samuel Neural Checkers,1959-07-01,428400000.0,1
Pandemonium (morse),1959-02-01,600000000.0,1
Perceptron Mark I,1957-01-01,694894.9377361819,1
Theseus,1950-07-02,40.0,1