{
 "run": "586-chinchilla-scaling-break",
 "as_of": "2026-10-30",
 "source": "Epoch AI, notable_ai_models.csv (epoch.ai/data), keyless CSV export of their public Notable AI Models dataset -- Domain == 'Language', Base model empty (drops fine-tunes, whose reported training-dataset size is the fine-tune set, not the pretraining corpus), Parameters and Training dataset size (total) both present, Publication date >= 2018-01-01.",
 "n": 174,
 "cutoff_label": "Chinchilla paper (arXiv:2203.15556, 2022-03-29)",
 "full_fit": {
  "label": "Full sample, 2018-2026",
  "n": 174,
  "slope": 0.2702,
  "slope_ci95": [
   0.2003,
   0.34
  ],
  "r2": 0.253,
  "p": 1.511502179745615e-12
 },
 "pre_fit": {
  "label": "Before Chinchilla",
  "n": 100,
  "slope": 0.2919,
  "slope_ci95": [
   0.096,
   0.4879
  ],
  "r2": 0.0819,
  "p": 0.0039040010514192936
 },
 "post_fit": {
  "label": "After Chinchilla",
  "n": 74,
  "slope": 0.3704,
  "slope_ci95": [
   0.2039,
   0.5369
  ],
  "r2": 0.2146,
  "p": 3.2415398300018605e-05
 },
 "break_test": {
  "classical_se": {
   "coef": 0.0785,
   "ci95": [
    -0.1882,
    0.3452
   ],
   "p": 0.5620901514816605
  },
  "hc3_robust_se": {
   "coef": 0.0785,
   "ci95": [
    -0.1582,
    0.3151
   ],
   "p": 0.515690834206384
  }
 },
 "alt_breaks": [
  {
   "label": "ChatGPT release (2022-11-30)",
   "coef": 0.0378,
   "ci95": [
    -0.2741,
    0.3497
   ],
   "p": 0.8110968476006893
  },
  {
   "label": "Llama 1 release (2023-02-24)",
   "coef": 0.0211,
   "ci95": [
    -0.3064,
    0.3486
   ],
   "p": 0.8989452994088041
  }
 ],
 "confidence_filtered": {
  "full_fit": {
   "label": "Confident-only, full sample",
   "n": 119,
   "slope": 0.2935,
   "slope_ci95": [
    0.201,
    0.386
   ],
   "r2": 0.2525,
   "p": 5.797160725452475e-09
  },
  "n": 119,
  "break_test": {
   "coef": 0.1061,
   "ci95": [
    -0.2355,
    0.4477
   ],
   "p": 0.5394830905296606
  }
 },
 "outlier_dropped": {
  "n": 173,
  "break_test": {
   "coef": 0.0708,
   "ci95": [
    -0.1942,
    0.3359
   ],
   "p": 0.5985043661405915
  }
 },
 "snapshots": {
  "median_ratio_pre": 1.8,
  "median_ratio_post": 28.57,
  "mean_log_gap": 0.9659,
  "mean_log_gap_ci95": [
   0.6285,
   1.2865
  ],
  "frac_ge20_pre": 0.27,
  "frac_ge20_post": 0.6622
 },
 "table": [
  {
   "model": "QRNN",
   "organization": "Salesforce Research",
   "publication_date": "2018-02-01",
   "parameters": 135000000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.762962962962963,
   "confidence": "Likely"
  },
  {
   "model": "ELMo",
   "organization": "University of Washington,Allen Institute for AI",
   "publication_date": "2018-02-01",
   "parameters": 94000000.0,
   "training_tokens": 2000000000.0,
   "tokens_per_param": 21.27659574468085,
   "confidence": "Speculative"
  },
  {
   "model": "4 layer QRNN (h=2500)",
   "organization": "Salesforce Research",
   "publication_date": "2018-03-22",
   "parameters": 151000000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.6821192052980133,
   "confidence": "Likely"
  },
  {
   "model": "Dropout-LSTM+Noise(Bernoulli) (WT2)",
   "organization": "Columbia University,New York University (NYU),Princeton University",
   "publication_date": "2018-05-03",
   "parameters": 51000000.0,
   "training_tokens": 2000000.0,
   "tokens_per_param": 0.0392156862745098,
   "confidence": "Likely"
  },
  {
   "model": "aLSTM(depth-2)+RecurrentPolicy (WT2)",
   "organization": "University of Manchester,Alan Turing Institute",
   "publication_date": "2018-05-22",
   "parameters": 32000000.0,
   "training_tokens": 2000000.0,
   "tokens_per_param": 0.0625,
   "confidence": "Confident"
  },
  {
   "model": "GPT-1",
   "organization": "OpenAI",
   "publication_date": "2018-06-01",
   "parameters": 117000000.0,
   "training_tokens": 1333333333.0,
   "tokens_per_param": 11.396011393162391,
   "confidence": "Likely"
  },
  {
   "model": "AWD-LSTM-MoS+PDR + dynamic evaluation (WT2)",
   "organization": "IBM",
   "publication_date": "2018-08-14",
   "parameters": 35000000.0,
   "training_tokens": 2000000.0,
   "tokens_per_param": 0.0571428571428571,
   "confidence": "Confident"
  },
  {
   "model": "(ensemble): AWD-LSTM-DOC (fin) \u00d7 5 (WT2)",
   "organization": "NTT Communication Science Laboratories,Tohoku University",
   "publication_date": "2018-08-30",
   "parameters": 185000000.0,
   "training_tokens": 2000000.0,
   "tokens_per_param": 0.0108108108108108,
   "confidence": "Confident"
  },
  {
   "model": "Transformer + Simple Recurrent Unit",
   "organization": "ASAPP,Cornell University,Google,Princeton University",
   "publication_date": "2018-09-17",
   "parameters": 90000000.0,
   "training_tokens": 112500000.0,
   "tokens_per_param": 1.25,
   "confidence": "Confident"
  },
  {
   "model": "AWD-LSTM-MoS + dynamic evaluation (WT2, 2018)",
   "organization": "Peking University,Microsoft Research Asia",
   "publication_date": "2018-09-18",
   "parameters": 35000000.0,
   "training_tokens": 2000000.0,
   "tokens_per_param": 0.0571428571428571,
   "confidence": "Confident"
  },
  {
   "model": "LSTM+NeuralCache",
   "organization": "KU Leuven,ESAT - PSI,Apple",
   "publication_date": "2018-09-24",
   "parameters": 2100000.0,
   "training_tokens": 2000000.0,
   "tokens_per_param": 0.9523809523809524,
   "confidence": "Likely"
  },
  {
   "model": "Transformer (Adaptive Input Embeddings) WT103",
   "organization": "Facebook AI Research",
   "publication_date": "2018-09-28",
   "parameters": 247000000.0,
   "training_tokens": 100000000.0,
   "tokens_per_param": 0.4048582995951417,
   "confidence": "Confident"
  },
  {
   "model": "BERT-Large",
   "organization": "Google",
   "publication_date": "2018-10-11",
   "parameters": 340000000.0,
   "training_tokens": 2649900000.0,
   "tokens_per_param": 7.793823529411765,
   "confidence": "Confident"
  },
  {
   "model": "TrellisNet",
   "organization": "Carnegie Mellon University (CMU),Bosch Center for Artificial Intelligence,Intel Labs",
   "publication_date": "2018-10-15",
   "parameters": 180000000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.5722222222222222,
   "confidence": "Confident"
  },
  {
   "model": "Mesh-TensorFlow Transformer 4.9B (language)",
   "organization": "Google Brain",
   "publication_date": "2018-11-05",
   "parameters": 4900000000.0,
   "training_tokens": 5000000000.0,
   "tokens_per_param": 1.020408163265306,
   "confidence": "Confident"
  },
  {
   "model": "Mesh-TensorFlow Transformer 2.9B (translation)",
   "organization": "Google Brain",
   "publication_date": "2018-11-05",
   "parameters": 2900000000.0,
   "training_tokens": 1550000000.0,
   "tokens_per_param": 0.5344827586206896,
   "confidence": "Likely"
  },
  {
   "model": "Multi-cell LSTM",
   "organization": "University of Hyderabad",
   "publication_date": "2018-11-15",
   "parameters": 7200000.0,
   "training_tokens": 929000.0,
   "tokens_per_param": 0.1290277777777777,
   "confidence": "Likely"
  },
  {
   "model": "GPipe (Transformer)",
   "organization": "Google",
   "publication_date": "2018-11-16",
   "parameters": 6000000000.0,
   "training_tokens": 1450000000000.0,
   "tokens_per_param": 241.66666666666663,
   "confidence": "Likely"
  },
  {
   "model": "Transformer ELMo",
   "organization": "Allen Institute for AI,University of Washington",
   "publication_date": "2019-01-01",
   "parameters": 56000000.0,
   "training_tokens": 2000000000.0,
   "tokens_per_param": 35.714285714285715,
   "confidence": "Confident"
  },
  {
   "model": "Transformer-XL (257M)",
   "organization": "Carnegie Mellon University (CMU),Google Brain",
   "publication_date": "2019-01-09",
   "parameters": 257000000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.4007782101167315,
   "confidence": "Speculative"
  },
  {
   "model": "MT-DNN",
   "organization": "Microsoft",
   "publication_date": "2019-01-31",
   "parameters": 330000000.0,
   "training_tokens": 1000000.0,
   "tokens_per_param": 0.003030303030303,
   "confidence": "Confident"
  },
  {
   "model": "GPT-2 (1.5B)",
   "organization": "OpenAI",
   "publication_date": "2019-02-14",
   "parameters": 1500000000.0,
   "training_tokens": 10666666666.666666,
   "tokens_per_param": 7.111111111111111,
   "confidence": "Speculative"
  },
  {
   "model": "SciBERT",
   "organization": "Allen Institute for AI",
   "publication_date": "2019-03-26",
   "parameters": 110000000.0,
   "training_tokens": 3170000000.0,
   "tokens_per_param": 28.818181818181817,
   "confidence": "Confident"
  },
  {
   "model": "WeNet (Penn Treebank)",
   "organization": "Amazon",
   "publication_date": "2019-04-08",
   "parameters": 23000000.0,
   "training_tokens": 929000.0,
   "tokens_per_param": 0.040391304347826,
   "confidence": "Likely"
  },
  {
   "model": "True-Regularization+Finetune+Dynamic-Eval",
   "organization": "Mobvoi,Williams College",
   "publication_date": "2019-04-08",
   "parameters": 7000000.0,
   "training_tokens": 929000.0,
   "tokens_per_param": 0.1327142857142857,
   "confidence": "Confident"
  },
  {
   "model": "Transformer-XL + RMS dynamic eval",
   "organization": "University of Edinburgh",
   "publication_date": "2019-04-17",
   "parameters": 257000000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.4007782101167315,
   "confidence": "Confident"
  },
  {
   "model": "BERT-Large-CAS (PTB+WT2+WT103)",
   "organization": "Amazon",
   "publication_date": "2019-04-20",
   "parameters": 395000000.0,
   "training_tokens": 1300000000.0,
   "tokens_per_param": 3.2911392405063293,
   "confidence": "Likely"
  },
  {
   "model": "AWD-LSTM-DRILL + dynamic evaluation\u2020 (WT2)",
   "organization": "IDIAP",
   "publication_date": "2019-05-14",
   "parameters": 34000000.0,
   "training_tokens": 2000000.0,
   "tokens_per_param": 0.0588235294117647,
   "confidence": "Likely"
  },
  {
   "model": "XLNet",
   "organization": "Carnegie Mellon University (CMU),Google Brain",
   "publication_date": "2019-06-01",
   "parameters": 340000000.0,
   "training_tokens": 32890000000.0,
   "tokens_per_param": 96.73529411764706,
   "confidence": "Confident"
  },
  {
   "model": "AWD-LSTM + MoS + Partial Shuffled",
   "organization": "University of Texas at Austin",
   "publication_date": "2019-06-10",
   "parameters": 35000000.0,
   "training_tokens": 2000000.0,
   "tokens_per_param": 0.0571428571428571,
   "confidence": "Confident"
  },
  {
   "model": "RoBERTa Large",
   "organization": "Facebook,University of Washington",
   "publication_date": "2019-07-01",
   "parameters": 355000000.0,
   "training_tokens": 42666666666.0,
   "tokens_per_param": 120.18779342535213,
   "confidence": "Confident"
  },
  {
   "model": "Mogrifier (d2, MoS2, MC) + dynamic eval",
   "organization": "DeepMind,University of Oxford",
   "publication_date": "2019-09-04",
   "parameters": 35000000.0,
   "training_tokens": 2000000.0,
   "tokens_per_param": 0.0571428571428571,
   "confidence": "Confident"
  },
  {
   "model": "Megatron-LM (8.3B)",
   "organization": "Nvidia",
   "publication_date": "2019-09-17",
   "parameters": 8300000000.0,
   "training_tokens": 46400000000.0,
   "tokens_per_param": 5.590361445783133,
   "confidence": "Likely"
  },
  {
   "model": "Megatron-LM (1.2B)",
   "organization": "Nvidia",
   "publication_date": "2019-09-17",
   "parameters": 1200000000.0,
   "training_tokens": 157000000000.0,
   "tokens_per_param": 130.83333333333334,
   "confidence": "Confident"
  },
  {
   "model": "Megatron-BERT",
   "organization": "Nvidia",
   "publication_date": "2019-09-17",
   "parameters": 3900000000.0,
   "training_tokens": 6960000000.0,
   "tokens_per_param": 1.7846153846153847,
   "confidence": "Confident"
  },
  {
   "model": "Adaptive Inputs + LayerDrop",
   "organization": "Facebook AI Research,LORIA",
   "publication_date": "2019-09-25",
   "parameters": 423000000.00000006,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.243498817966903,
   "confidence": "Confident"
  },
  {
   "model": "ALBERT",
   "organization": "Toyota Technological Institute at Chicago,Google Research",
   "publication_date": "2019-09-26",
   "parameters": 18000000.0,
   "training_tokens": 3300000000.0,
   "tokens_per_param": 183.33333333333331,
   "confidence": "Confident"
  },
  {
   "model": "DistilBERT",
   "organization": "Hugging Face",
   "publication_date": "2019-10-02",
   "parameters": 66000000.0,
   "training_tokens": 495000000.0,
   "tokens_per_param": 7.5,
   "confidence": "Confident"
  },
  {
   "model": "T5-11B",
   "organization": "Google",
   "publication_date": "2019-10-23",
   "parameters": 11000000000.0,
   "training_tokens": 34000000000.0,
   "tokens_per_param": 3.090909090909091,
   "confidence": "Confident"
  },
  {
   "model": "T5-3B",
   "organization": "Google",
   "publication_date": "2019-10-23",
   "parameters": 2800000000.0,
   "training_tokens": 5100000000.0,
   "tokens_per_param": 1.8214285714285716,
   "confidence": "Confident"
  },
  {
   "model": "BART-large",
   "organization": "Facebook AI",
   "publication_date": "2019-10-29",
   "parameters": 406291456.0,
   "training_tokens": 42666666666.0,
   "tokens_per_param": 105.01492471946052,
   "confidence": "Likely"
  },
  {
   "model": "Base LM + kNN LM + Continuous Cache",
   "organization": "Stanford University,Facebook AI Research",
   "publication_date": "2019-11-01",
   "parameters": 247000000.00000003,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.4170040485829959,
   "confidence": "Likely"
  },
  {
   "model": "XLM-RoBERTa",
   "organization": "Facebook AI",
   "publication_date": "2019-11-05",
   "parameters": 550000000.0,
   "training_tokens": 167000000000.0,
   "tokens_per_param": 303.6363636363636,
   "confidence": "Confident"
  },
  {
   "model": "Sandwich Transformer",
   "organization": "Allen Institute for AI,Facebook AI Research",
   "publication_date": "2019-11-10",
   "parameters": 209000000.0,
   "training_tokens": 700000000.0,
   "tokens_per_param": 3.349282296650718,
   "confidence": "Speculative"
  },
  {
   "model": "CamemBERT",
   "organization": "Facebook,INRIA,Sorbonne University",
   "publication_date": "2019-11-10",
   "parameters": 335000000.0,
   "training_tokens": 28615771779.58404,
   "tokens_per_param": 85.42021426741505,
   "confidence": "Confident"
  },
  {
   "model": "Transformer-XL DeFINE (141M)",
   "organization": "University of Washington,Allen Institute for AI",
   "publication_date": "2019-11-27",
   "parameters": 141000000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.7304964539007093,
   "confidence": "Speculative"
  },
  {
   "model": "MMLSTM (WT-2)",
   "organization": "Beijing University of Posts and Telecommunications,University of West London",
   "publication_date": "2019-12-05",
   "parameters": 32300000.0,
   "training_tokens": 2000000.0,
   "tokens_per_param": 0.0619195046439628,
   "confidence": "Likely"
  },
  {
   "model": "MMLSTM (PTB)",
   "organization": "Beijing University of Posts and Telecommunications,University of West London",
   "publication_date": "2019-12-05",
   "parameters": 21300000.0,
   "training_tokens": 929000.0,
   "tokens_per_param": 0.0436150234741784,
   "confidence": "Likely"
  },
  {
   "model": "Meena",
   "organization": "Google Brain",
   "publication_date": "2020-01-28",
   "parameters": 2600000000.0,
   "training_tokens": 53333333333.333336,
   "tokens_per_param": 20.512820512820515,
   "confidence": "Confident"
  },
  {
   "model": "TaLK Convolution",
   "organization": "Carleton University",
   "publication_date": "2020-02-08",
   "parameters": 240000000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.4291666666666666,
   "confidence": "Likely"
  },
  {
   "model": "ALBERT-xxlarge",
   "organization": "Toyota Technological Institute at Chicago,Google",
   "publication_date": "2020-02-09",
   "parameters": 235000000.0,
   "training_tokens": 3300000000.0,
   "tokens_per_param": 14.04255319148936,
   "confidence": "Speculative"
  },
  {
   "model": "Turing-NLG",
   "organization": "Microsoft",
   "publication_date": "2020-02-13",
   "parameters": 17000000000.0,
   "training_tokens": 46400000000.0,
   "tokens_per_param": 2.7294117647058824,
   "confidence": "Likely"
  },
  {
   "model": "FFN SwiGLU",
   "organization": "Google",
   "publication_date": "2020-02-14",
   "parameters": 220000000.0,
   "training_tokens": 50600083456.0,
   "tokens_per_param": 230.00037934545452,
   "confidence": "Confident"
  },
  {
   "model": "Feedback Transformer",
   "organization": "LORIA,University of Lorraine,Facebook AI Research",
   "publication_date": "2020-02-21",
   "parameters": 126000000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.8174603174603174,
   "confidence": "Speculative"
  },
  {
   "model": "TCAN (WT2)",
   "organization": "Nanjing University,Ant Group",
   "publication_date": "2020-02-28",
   "parameters": 33000000.0,
   "training_tokens": 2000000.0,
   "tokens_per_param": 0.0606060606060606,
   "confidence": "Confident"
  },
  {
   "model": "TransformerXL + spectrum control",
   "organization": "University of California Los Angeles (UCLA),JD.com",
   "publication_date": "2020-03-11",
   "parameters": 151000000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.6821192052980133,
   "confidence": "Speculative"
  },
  {
   "model": "Routing Transformer (WT-103)",
   "organization": "Google Research",
   "publication_date": "2020-03-12",
   "parameters": 79500000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 1.2955974842767295,
   "confidence": "Confident"
  },
  {
   "model": "Tensor-Transformer(1core)+PN (WT103)",
   "organization": "University of California (UC) Berkeley",
   "publication_date": "2020-03-17",
   "parameters": 85300000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 1.2075029308323564,
   "confidence": "Confident"
  },
  {
   "model": "ELECTRA",
   "organization": "Stanford University,Google,Google Brain",
   "publication_date": "2020-03-23",
   "parameters": 335000000.0,
   "training_tokens": 33000000000.0,
   "tokens_per_param": 98.50746268656717,
   "confidence": "Likely"
  },
  {
   "model": "Retrieval-Augmented Generator",
   "organization": "Facebook,New York University (NYU),University College London (UCL)",
   "publication_date": "2020-05-22",
   "parameters": 626000000.0,
   "training_tokens": 3074560.0,
   "tokens_per_param": 0.0049114376996805,
   "confidence": "Confident"
  },
  {
   "model": "GPT-3 175B (davinci)",
   "organization": "OpenAI",
   "publication_date": "2020-05-28",
   "parameters": 174600000000.0,
   "training_tokens": 238000000000.0,
   "tokens_per_param": 1.3631156930126005,
   "confidence": "Confident"
  },
  {
   "model": "GShard (dense)",
   "organization": "Google",
   "publication_date": "2020-06-30",
   "parameters": 2300000000.0,
   "training_tokens": 346666666667.0,
   "tokens_per_param": 150.72463768130436,
   "confidence": "Confident"
  },
  {
   "model": "DeLighT",
   "organization": "University of Washington,Allen Institute for AI,Facebook AI Research",
   "publication_date": "2020-08-03",
   "parameters": 99000000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 1.0404040404040404,
   "confidence": "Likely"
  },
  {
   "model": "ERNIE-GEN (large)",
   "organization": "Baidu",
   "publication_date": "2020-08-06",
   "parameters": 340000000.0,
   "training_tokens": 114666666666.66669,
   "tokens_per_param": 337.2549019607843,
   "confidence": "Speculative"
  },
  {
   "model": "mT5-XXL",
   "organization": "Google,Google Research",
   "publication_date": "2020-10-20",
   "parameters": 13000000000.0,
   "training_tokens": 1000000000000.0,
   "tokens_per_param": 76.92307692307692,
   "confidence": "Confident"
  },
  {
   "model": "German ELECTRA Large",
   "organization": "deepset,Bayerische Staatsbibliothek Muenchen",
   "publication_date": "2020-10-21",
   "parameters": 335000000.0,
   "training_tokens": 36383733333.0,
   "tokens_per_param": 108.60815920298508,
   "confidence": "Confident"
  },
  {
   "model": "CPM-Large",
   "organization": "Tsinghua University,Beijing Academy of Artificial Intelligence / BAAI",
   "publication_date": "2020-12-01",
   "parameters": 2600000000.0,
   "training_tokens": 16700000000.0,
   "tokens_per_param": 6.423076923076923,
   "confidence": "Confident"
  },
  {
   "model": "CT-MoS (WT2)",
   "organization": "Google,National Tsing Hua University",
   "publication_date": "2020-12-25",
   "parameters": 45000000.0,
   "training_tokens": 2000000.0,
   "tokens_per_param": 0.0444444444444444,
   "confidence": "Confident"
  },
  {
   "model": "ERNIE-Doc (247M)",
   "organization": "Baidu",
   "publication_date": "2020-12-31",
   "parameters": 247000000.00000003,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.4170040485829959,
   "confidence": "Confident"
  },
  {
   "model": "Switch",
   "organization": "Google",
   "publication_date": "2021-01-11",
   "parameters": 1571000000000.0,
   "training_tokens": 86400000000.0,
   "tokens_per_param": 0.0549968173138128,
   "confidence": "Confident"
  },
  {
   "model": "SRU++ Large",
   "organization": "ASAPP",
   "publication_date": "2021-02-24",
   "parameters": 234000000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.4401709401709401,
   "confidence": "Confident"
  },
  {
   "model": "Generative BST",
   "organization": "Facebook AI Research",
   "publication_date": "2021-03-05",
   "parameters": 9431810048.0,
   "training_tokens": 56800000000.0,
   "tokens_per_param": 6.022173868105449,
   "confidence": "Confident"
  },
  {
   "model": "PLUG",
   "organization": "Alibaba",
   "publication_date": "2021-04-19",
   "parameters": 27000000000.0,
   "training_tokens": 60000000000.0,
   "tokens_per_param": 2.2222222222222223,
   "confidence": "Confident"
  },
  {
   "model": "ByT5-XXL",
   "organization": "Google,Google Research",
   "publication_date": "2021-05-28",
   "parameters": 12900000000.0,
   "training_tokens": 1048576000000.0,
   "tokens_per_param": 81.28496124031008,
   "confidence": "Likely"
  },
  {
   "model": "EMDR",
   "organization": "Mila - Quebec AI (originally Montreal Institute for Learning Algorithms),McGill University,DeepMind",
   "publication_date": "2021-06-09",
   "parameters": 440000000.0,
   "training_tokens": 171600000000.0,
   "tokens_per_param": 390.0,
   "confidence": "Confident"
  },
  {
   "model": "DeBERTa",
   "organization": "Microsoft",
   "publication_date": "2021-06-10",
   "parameters": 1500000000.0,
   "training_tokens": 20800000000.0,
   "tokens_per_param": 13.866666666666667,
   "confidence": "Confident"
  },
  {
   "model": "Adaptive Input Transformer + RD",
   "organization": "Microsoft Research Asia,Soochow University",
   "publication_date": "2021-06-28",
   "parameters": 247000000.00000003,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.4170040485829959,
   "confidence": "Likely"
  },
  {
   "model": "ERNIE 3.0",
   "organization": "Baidu",
   "publication_date": "2021-07-05",
   "parameters": 10000000000.0,
   "training_tokens": 375000000000.0,
   "tokens_per_param": 37.5,
   "confidence": "Confident"
  },
  {
   "model": "Jurassic-1-Jumbo",
   "organization": "AI21 Labs",
   "publication_date": "2021-08-11",
   "parameters": 178000000000.0,
   "training_tokens": 300000000000.0,
   "tokens_per_param": 1.6853932584269662,
   "confidence": "Confident"
  },
  {
   "model": "XLMR-XXL",
   "organization": "Facebook AI Research",
   "publication_date": "2021-08-17",
   "parameters": 10700000000.0,
   "training_tokens": 167000000000.0,
   "tokens_per_param": 15.607476635514018,
   "confidence": "Confident"
  },
  {
   "model": "PermuteFormer",
   "organization": "Peking University",
   "publication_date": "2021-09-06",
   "parameters": 149697024.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.6880564305673839,
   "confidence": "Speculative"
  },
  {
   "model": "HyperCLOVA 204B",
   "organization": "NAVER",
   "publication_date": "2021-09-10",
   "parameters": 204000000000.0,
   "training_tokens": 560000000000.0,
   "tokens_per_param": 2.7450980392156863,
   "confidence": "Speculative"
  },
  {
   "model": "PLATO-XL",
   "organization": "Baidu",
   "publication_date": "2021-09-20",
   "parameters": 11000000000.0,
   "training_tokens": 150000000000.0,
   "tokens_per_param": 13.636363636363637,
   "confidence": "Confident"
  },
  {
   "model": "Megatron-Turing NLG 530B",
   "organization": "Microsoft,Nvidia",
   "publication_date": "2021-10-11",
   "parameters": 530000000000.0,
   "training_tokens": 270000000000.0,
   "tokens_per_param": 0.5094339622641509,
   "confidence": "Confident"
  },
  {
   "model": "Yuan 1.0",
   "organization": "Inspur",
   "publication_date": "2021-10-12",
   "parameters": 245730000000.0,
   "training_tokens": 180000000000.0,
   "tokens_per_param": 0.7325112928824319,
   "confidence": "Confident"
  },
  {
   "model": "S4",
   "organization": "Stanford University",
   "publication_date": "2021-10-31",
   "parameters": 249000000.00000003,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.4136546184738955,
   "confidence": "Likely"
  },
  {
   "model": "Gopher (280B)",
   "organization": "DeepMind",
   "publication_date": "2021-12-08",
   "parameters": 280000000000.0,
   "training_tokens": 300000000000.0,
   "tokens_per_param": 1.0714285714285714,
   "confidence": "Confident"
  },
  {
   "model": "GLaM",
   "organization": "Google",
   "publication_date": "2021-12-13",
   "parameters": 1200000000000.0,
   "training_tokens": 600000000000.0,
   "tokens_per_param": 0.5,
   "confidence": "Confident"
  },
  {
   "model": "LongT5",
   "organization": "Google Research",
   "publication_date": "2021-12-15",
   "parameters": 3000000000.0,
   "training_tokens": 524288000000.0,
   "tokens_per_param": 174.76266666666666,
   "confidence": "Confident"
  },
  {
   "model": "XGLM-7.5B",
   "organization": "Meta AI,Facebook AI Research",
   "publication_date": "2021-12-20",
   "parameters": 7500000000.0,
   "training_tokens": 500000000000.0,
   "tokens_per_param": 66.66666666666667,
   "confidence": "Confident"
  },
  {
   "model": "ERNIE 3.0 Titan",
   "organization": "Baidu,Peng Cheng Laboratory",
   "publication_date": "2021-12-23",
   "parameters": 260000000000.0,
   "training_tokens": 668000000000.0,
   "tokens_per_param": 2.569230769230769,
   "confidence": "Confident"
  },
  {
   "model": "data2vec (language)",
   "organization": "Meta AI",
   "publication_date": "2022-01-20",
   "parameters": 705134592.0,
   "training_tokens": 131072000000.0,
   "tokens_per_param": 185.88224359868025,
   "confidence": "Likely"
  },
  {
   "model": "AlphaCode",
   "organization": "DeepMind",
   "publication_date": "2022-02-02",
   "parameters": 41100000000.0,
   "training_tokens": 967000000000.0,
   "tokens_per_param": 23.52798053527981,
   "confidence": "Confident"
  },
  {
   "model": "RETRO-7B",
   "organization": "DeepMind",
   "publication_date": "2022-02-07",
   "parameters": 7500000000.0,
   "training_tokens": 419430400000.0,
   "tokens_per_param": 55.92405333333333,
   "confidence": "Confident"
  },
  {
   "model": "GPT-NeoX-20B",
   "organization": "EleutherAI",
   "publication_date": "2022-02-09",
   "parameters": 20000000000.0,
   "training_tokens": 341173367965.0,
   "tokens_per_param": 17.05866839825,
   "confidence": "Confident"
  },
  {
   "model": "LaMDA",
   "organization": "Google",
   "publication_date": "2022-02-10",
   "parameters": 137000000000.0,
   "training_tokens": 2080000000000.0,
   "tokens_per_param": 15.182481751824817,
   "confidence": "Confident"
  },
  {
   "model": "ST-MoE",
   "organization": "Google,Google Brain,Google Research",
   "publication_date": "2022-02-17",
   "parameters": 269000000000.0,
   "training_tokens": 1500000000000.0,
   "tokens_per_param": 5.5762081784386615,
   "confidence": "Likely"
  },
  {
   "model": "PolyCoder",
   "organization": "Carnegie Mellon University (CMU)",
   "publication_date": "2022-02-26",
   "parameters": 2700000000.0,
   "training_tokens": 39300000000.0,
   "tokens_per_param": 14.555555555555555,
   "confidence": "Likely"
  },
  {
   "model": "DeepNet",
   "organization": "Microsoft Research",
   "publication_date": "2022-03-01",
   "parameters": 3200000000.0,
   "training_tokens": 272629760000.0,
   "tokens_per_param": 85.1968,
   "confidence": "Confident"
  },
  {
   "model": "Segatron-XL large, M=384 + HCP",
   "organization": "Microsoft Research,University of Waterloo",
   "publication_date": "2022-03-21",
   "parameters": 257000000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.4007782101167315,
   "confidence": "Confident"
  },
  {
   "model": "Chinchilla",
   "organization": "DeepMind",
   "publication_date": "2022-03-29",
   "parameters": 70000000000.0,
   "training_tokens": 1400000000000.0,
   "tokens_per_param": 20.0,
   "confidence": "Confident"
  },
  {
   "model": "PaLM (540B)",
   "organization": "Google Research",
   "publication_date": "2022-04-04",
   "parameters": 540350000000.0,
   "training_tokens": 780000000000.0,
   "tokens_per_param": 1.4435088368649949,
   "confidence": "Confident"
  },
  {
   "model": "Sparse all-MLP",
   "organization": "Meta AI",
   "publication_date": "2022-04-14",
   "parameters": 9410000000.0,
   "training_tokens": 100000000000.0,
   "tokens_per_param": 10.626992561105208,
   "confidence": "Confident"
  },
  {
   "model": "OPT-175B",
   "organization": "Meta AI",
   "publication_date": "2022-05-02",
   "parameters": 175000000000.0,
   "training_tokens": 180000000000.0,
   "tokens_per_param": 1.0285714285714285,
   "confidence": "Confident"
  },
  {
   "model": "UL2",
   "organization": "Google Research,Google Brain",
   "publication_date": "2022-05-10",
   "parameters": 20000000000.0,
   "training_tokens": 1000000000000.0,
   "tokens_per_param": 50.0,
   "confidence": "Confident"
  },
  {
   "model": "GPT-2 Medium (FlashAttention)",
   "organization": "Stanford University,University at Buffalo",
   "publication_date": "2022-05-27",
   "parameters": 355000000.0,
   "training_tokens": 10130000000.0,
   "tokens_per_param": 28.535211267605632,
   "confidence": "Confident"
  },
  {
   "model": "DITTO",
   "organization": "Tsinghua University,Apple,Westlake University,Chinese University of Hong Kong (CUHK)",
   "publication_date": "2022-06-06",
   "parameters": 750000000.0,
   "training_tokens": 103000000.0,
   "tokens_per_param": 0.1373333333333333,
   "confidence": "Confident"
  },
  {
   "model": "CodeT5-large",
   "organization": "Salesforce",
   "publication_date": "2022-07-05",
   "parameters": 770000000.0,
   "training_tokens": 10500000000.0,
   "tokens_per_param": 13.636363636363637,
   "confidence": "Likely"
  },
  {
   "model": "NLLB",
   "organization": "Meta AI",
   "publication_date": "2022-07-06",
   "parameters": 54500000000.0,
   "training_tokens": 300000000000.0,
   "tokens_per_param": 5.504587155963303,
   "confidence": "Confident"
  },
  {
   "model": "BLOOM-176B",
   "organization": "Hugging Face,BigScience",
   "publication_date": "2022-07-11",
   "parameters": 176247271424.0,
   "training_tokens": 379000000000.0,
   "tokens_per_param": 2.150387900691157,
   "confidence": "Confident"
  },
  {
   "model": "AlexaTM 20B",
   "organization": "Amazon",
   "publication_date": "2022-08-02",
   "parameters": 19750000000.0,
   "training_tokens": 1319000000000.0,
   "tokens_per_param": 66.78481012658227,
   "confidence": "Confident"
  },
  {
   "model": "GLM-130B",
   "organization": "Tsinghua University",
   "publication_date": "2022-08-04",
   "parameters": 130000000000.0,
   "training_tokens": 152000000000.0,
   "tokens_per_param": 1.1692307692307693,
   "confidence": "Confident"
  },
  {
   "model": "Mogrifier RLSTM (WT2)",
   "organization": "DeepMind",
   "publication_date": "2022-11-03",
   "parameters": 35000000.0,
   "training_tokens": 2666667.0,
   "tokens_per_param": 0.0761904857142857,
   "confidence": "Confident"
  },
  {
   "model": "Fusion in Encoder",
   "organization": "Samsung",
   "publication_date": "2022-11-18",
   "parameters": 330000000.0,
   "training_tokens": 960000.0,
   "tokens_per_param": 0.0029090909090909,
   "confidence": "Likely"
  },
  {
   "model": "ALM 1.0",
   "organization": "Beijing Academy of Artificial Intelligence / BAAI",
   "publication_date": "2022-11-28",
   "parameters": 335000000.0,
   "training_tokens": 22696572400.0,
   "tokens_per_param": 67.7509623880597,
   "confidence": "Speculative"
  },
  {
   "model": "Vega v2",
   "organization": "Wuhan University,JD Explore Academy,Shanghai AI Lab,Nanyang Technological University,Washington University in St Louis,Chongqing University of Posts and Telecommunications,University of Sydney",
   "publication_date": "2022-12-04",
   "parameters": 6000000000.0,
   "training_tokens": 6439999999.0,
   "tokens_per_param": 1.0733333331666666,
   "confidence": "Confident"
  },
  {
   "model": "Hybrid H3-2.7B",
   "organization": "Stanford University,University at Buffalo",
   "publication_date": "2022-12-28",
   "parameters": 2700000000.0,
   "training_tokens": 400000000000.0,
   "tokens_per_param": 148.14814814814815,
   "confidence": "Likely"
  },
  {
   "model": "LLaMA-65B",
   "organization": "Meta AI",
   "publication_date": "2023-02-24",
   "parameters": 65200000000.0,
   "training_tokens": 1400000000000.0,
   "tokens_per_param": 21.47239263803681,
   "confidence": "Confident"
  },
  {
   "model": "Falcon-40B",
   "organization": "Technology Innovation Institute",
   "publication_date": "2023-03-15",
   "parameters": 40000000000.0,
   "training_tokens": 1000000000000.0,
   "tokens_per_param": 25.0,
   "confidence": "Confident"
  },
  {
   "model": "PanGu-\u03a3",
   "organization": "Huawei Noah's Ark Lab",
   "publication_date": "2023-03-20",
   "parameters": 1085000000000.0,
   "training_tokens": 329000000000.0,
   "tokens_per_param": 0.3032258064516129,
   "confidence": "Confident"
  },
  {
   "model": "BloombergGPT",
   "organization": "Bloomberg,Johns Hopkins University",
   "publication_date": "2023-03-30",
   "parameters": 50558868480.0,
   "training_tokens": 569000000000.0,
   "tokens_per_param": 11.254207562518614,
   "confidence": "Confident"
  },
  {
   "model": "Incoder-6.7B",
   "organization": "Facebook AI Research,University of Washington,University of California (UC) Berkeley,Carnegie Mellon University (CMU),Toyota Technological Institute at Chicago",
   "publication_date": "2023-04-09",
   "parameters": 6700000000.0,
   "training_tokens": 52000000000.0,
   "tokens_per_param": 7.761194029850746,
   "confidence": "Confident"
  },
  {
   "model": "StarCoder",
   "organization": "Hugging Face,ServiceNow,Northeastern University,Mila - Quebec AI (originally Montreal Institute for Learning Algorithms),Carnegie Mellon University (CMU),Johns Hopkins University,Leipzig University,ScaDS.AI,Queen Mary University of London,Roblox,Sea AI Lab,Technion - Israel Institute of Technology,Monash University,CSIRO,Data61,McGill University,Saama,University of British Columbia (UBC),Massachusetts Institute of Technology (MIT),Technical University of Munich,IBM,University of Vermont,UnfoldML,SAP,University of Notre Dame,Columbia University,New York University (NYU),University of Allahabad,Discover Dollar,Toloka,Telefonica,Stanford University,Weizmann Institute of Science,Alan Turing Institute,Wellesley College,EleutherAI,Forschungszentrum Julich",
   "publication_date": "2023-05-09",
   "parameters": 15500000000.0,
   "training_tokens": 203750000000.0,
   "tokens_per_param": 13.14516129032258,
   "confidence": "Confident"
  },
  {
   "model": "PaLM 2",
   "organization": "Google",
   "publication_date": "2023-05-10",
   "parameters": 340000000000.0,
   "training_tokens": 3600000000000.0,
   "tokens_per_param": 10.588235294117649,
   "confidence": "Likely"
  },
  {
   "model": "CodeT5+",
   "organization": "Salesforce",
   "publication_date": "2023-05-20",
   "parameters": 16000000000.0,
   "training_tokens": 51500000000.0,
   "tokens_per_param": 3.21875,
   "confidence": "Confident"
  },
  {
   "model": "InternLM",
   "organization": "Shanghai AI Lab,SenseTime",
   "publication_date": "2023-07-06",
   "parameters": 104000000000.0,
   "training_tokens": 1600000000000.0,
   "tokens_per_param": 15.384615384615383,
   "confidence": "Confident"
  },
  {
   "model": "Llama 2-70B",
   "organization": "Meta AI",
   "publication_date": "2023-07-18",
   "parameters": 70000000000.0,
   "training_tokens": 2000000000000.0,
   "tokens_per_param": 28.571428571428573,
   "confidence": "Confident"
  },
  {
   "model": "Llama 2-7B",
   "organization": "Meta AI",
   "publication_date": "2023-07-18",
   "parameters": 7000000000.0,
   "training_tokens": 2000000000000.0,
   "tokens_per_param": 285.7142857142857,
   "confidence": "Confident"
  },
  {
   "model": "Jais",
   "organization": "Cerebras Systems,Mohamed bin Zayed University of Artificial Intelligence (MBZUAI),Inception G42",
   "publication_date": "2023-08-29",
   "parameters": 13000000000.0,
   "training_tokens": 395000000000.0,
   "tokens_per_param": 30.384615384615383,
   "confidence": "Confident"
  },
  {
   "model": "Falcon-180B",
   "organization": "Technology Innovation Institute",
   "publication_date": "2023-09-06",
   "parameters": 180000000000.0,
   "training_tokens": 3500000000000.0,
   "tokens_per_param": 19.444444444444443,
   "confidence": "Confident"
  },
  {
   "model": "CODEFUSION (Python)",
   "organization": "Microsoft,Microsoft Research",
   "publication_date": "2023-10-26",
   "parameters": 75000000.0,
   "training_tokens": 4390400.0,
   "tokens_per_param": 0.0585386666666666,
   "confidence": "Confident"
  },
  {
   "model": "Skywork-13B",
   "organization": "Kunlun Inc.",
   "publication_date": "2023-10-30",
   "parameters": 13000000000.0,
   "training_tokens": 3200000000000.0,
   "tokens_per_param": 246.15384615384616,
   "confidence": "Confident"
  },
  {
   "model": "Yi-34B",
   "organization": "01.AI",
   "publication_date": "2023-11-02",
   "parameters": 34000000000.0,
   "training_tokens": 3100000000000.0,
   "tokens_per_param": 91.17647058823528,
   "confidence": "Confident"
  },
  {
   "model": "Grok-1",
   "organization": "xAI",
   "publication_date": "2023-11-04",
   "parameters": 314000000000.0,
   "training_tokens": 6200000000000.0,
   "tokens_per_param": 19.745222929936304,
   "confidence": "Likely"
  },
  {
   "model": "RoFormer",
   "organization": "Zhuiyi Technology",
   "publication_date": "2023-11-08",
   "parameters": 110000000.0,
   "training_tokens": 3276800000.0,
   "tokens_per_param": 29.78909090909091,
   "confidence": "Confident"
  },
  {
   "model": "Nemotron-3-8B",
   "organization": "Nvidia",
   "publication_date": "2023-11-15",
   "parameters": 8000000000.0,
   "training_tokens": 3800000000000.0,
   "tokens_per_param": 475.0,
   "confidence": "Confident"
  },
  {
   "model": "Qwen-72B",
   "organization": "Alibaba",
   "publication_date": "2023-11-30",
   "parameters": 72000000000.0,
   "training_tokens": 3000000000000.0,
   "tokens_per_param": 41.66666666666666,
   "confidence": "Confident"
  },
  {
   "model": "Qwen1.5-72B",
   "organization": "Alibaba",
   "publication_date": "2024-02-04",
   "parameters": 72000000000.0,
   "training_tokens": 3000000000000.0,
   "tokens_per_param": 41.66666666666666,
   "confidence": "Confident"
  },
  {
   "model": "Aramco Metabrain AI",
   "organization": "Saudi Aramco",
   "publication_date": "2024-03-04",
   "parameters": 250000000000.0,
   "training_tokens": 7000000000000.0,
   "tokens_per_param": 28.0,
   "confidence": "Likely"
  },
  {
   "model": "DBRX",
   "organization": "Databricks",
   "publication_date": "2024-03-27",
   "parameters": 132000000000.0,
   "training_tokens": 12000000000000.0,
   "tokens_per_param": 90.9090909090909,
   "confidence": "Confident"
  },
  {
   "model": "Llama 3-70B",
   "organization": "Meta AI",
   "publication_date": "2024-04-18",
   "parameters": 70000000000.0,
   "training_tokens": 15000000000000.0,
   "tokens_per_param": 214.28571428571428,
   "confidence": "Confident"
  },
  {
   "model": "Yi-Large",
   "organization": "01.AI",
   "publication_date": "2024-05-13",
   "parameters": 100000000000.0,
   "training_tokens": 3000000000000.0,
   "tokens_per_param": 30.0,
   "confidence": "Speculative"
  },
  {
   "model": "Qwen2-72B",
   "organization": "Alibaba",
   "publication_date": "2024-06-07",
   "parameters": 72710000000.0,
   "training_tokens": 7000000000000.0,
   "tokens_per_param": 96.27286480539128,
   "confidence": "Confident"
  },
  {
   "model": "Nemotron-4 340B",
   "organization": "Nvidia",
   "publication_date": "2024-06-14",
   "parameters": 340000000000.0,
   "training_tokens": 9000000000000.0,
   "tokens_per_param": 26.47058823529412,
   "confidence": "Confident"
  },
  {
   "model": "Llama 3.1-405B",
   "organization": "Meta AI",
   "publication_date": "2024-07-23",
   "parameters": 405000000000.0,
   "training_tokens": 15600000000000.0,
   "tokens_per_param": 38.51851851851852,
   "confidence": "Confident"
  },
  {
   "model": "AFM-on-device",
   "organization": "Apple",
   "publication_date": "2024-07-29",
   "parameters": 2730000000.0,
   "training_tokens": 7588000000000.0,
   "tokens_per_param": 2779.48717948718,
   "confidence": "Confident"
  },
  {
   "model": "Qwen2.5-32B",
   "organization": "Alibaba",
   "publication_date": "2024-09-17",
   "parameters": 32500000000.0,
   "training_tokens": 18000000000000.0,
   "tokens_per_param": 553.8461538461538,
   "confidence": "Confident"
  },
  {
   "model": "Qwen2.5-72B",
   "organization": "Alibaba",
   "publication_date": "2024-09-19",
   "parameters": 72700000000.0,
   "training_tokens": 18000000000000.0,
   "tokens_per_param": 247.59284731774412,
   "confidence": "Confident"
  },
  {
   "model": "Telechat2-115B",
   "organization": "China Telecom",
   "publication_date": "2024-09-20",
   "parameters": 115000000000.0,
   "training_tokens": 10000000000000.0,
   "tokens_per_param": 86.95652173913044,
   "confidence": "Confident"
  },
  {
   "model": "Doubao-pro",
   "organization": "ByteDance",
   "publication_date": "2024-10-28",
   "parameters": 500000000000.0,
   "training_tokens": 8350000000000.0,
   "tokens_per_param": 16.7,
   "confidence": "Likely"
  },
  {
   "model": "Hunyuan-Large",
   "organization": "Tencent",
   "publication_date": "2024-11-06",
   "parameters": 389000000000.0,
   "training_tokens": 7000000000000.0,
   "tokens_per_param": 17.994858611825194,
   "confidence": "Confident"
  },
  {
   "model": "Llama 3.3 70B",
   "organization": "Meta AI",
   "publication_date": "2024-12-06",
   "parameters": 70000000000.0,
   "training_tokens": 15000000000000.0,
   "tokens_per_param": 214.28571428571428,
   "confidence": "Confident"
  },
  {
   "model": "EXAONE 3.5 32B",
   "organization": "LG AI Research",
   "publication_date": "2024-12-09",
   "parameters": 32000000000.0,
   "training_tokens": 6500000000000.0,
   "tokens_per_param": 203.125,
   "confidence": "Confident"
  },
  {
   "model": "DeepSeek-V3",
   "organization": "DeepSeek",
   "publication_date": "2024-12-24",
   "parameters": 671000000000.0,
   "training_tokens": 14800000000000.0,
   "tokens_per_param": 22.056631892697467,
   "confidence": "Confident"
  },
  {
   "model": "Hunyuan-TurboS",
   "organization": "Tencent",
   "publication_date": "2025-03-11",
   "parameters": 560000000000.0,
   "training_tokens": 16000000000000.0,
   "tokens_per_param": 28.571428571428573,
   "confidence": "Confident"
  },
  {
   "model": "DeepSeek-V3 (Mar 2025)",
   "organization": "DeepSeek",
   "publication_date": "2025-03-24",
   "parameters": 671000000000.0,
   "training_tokens": 14800000000000.0,
   "tokens_per_param": 22.056631892697467,
   "confidence": "Confident"
  },
  {
   "model": "Pangu Ultra",
   "organization": "Huawei",
   "publication_date": "2025-04-10",
   "parameters": 135000000000.0,
   "training_tokens": 13200000000000.0,
   "tokens_per_param": 97.77777777777776,
   "confidence": "Confident"
  },
  {
   "model": "Qwen3-235B-A22B",
   "organization": "Alibaba",
   "publication_date": "2025-04-29",
   "parameters": 235000000000.0,
   "training_tokens": 36000000000000.0,
   "tokens_per_param": 153.19148936170214,
   "confidence": "Likely"
  },
  {
   "model": "Kimi K2",
   "organization": "Moonshot",
   "publication_date": "2025-07-11",
   "parameters": 1000000000000.0,
   "training_tokens": 15500000000000.0,
   "tokens_per_param": 15.5,
   "confidence": "Confident"
  },
  {
   "model": "EXAONE 4.0 (32B)",
   "organization": "LG AI Research",
   "publication_date": "2025-07-15",
   "parameters": 32000000000.0,
   "training_tokens": 14000000000000.0,
   "tokens_per_param": 437.5,
   "confidence": "Confident"
  },
  {
   "model": "Qwen3-Coder-480B-A35B",
   "organization": "Alibaba",
   "publication_date": "2025-07-22",
   "parameters": 480000000000.0,
   "training_tokens": 7500000000000.0,
   "tokens_per_param": 15.625,
   "confidence": "Confident"
  },
  {
   "model": "Qwen3-235B-A22B-Thinking (Jul 2025)",
   "organization": "Alibaba",
   "publication_date": "2025-07-25",
   "parameters": 235000000000.0,
   "training_tokens": 36000000000000.0,
   "tokens_per_param": 153.19148936170214,
   "confidence": "Likely"
  },
  {
   "model": "Qwen3-235B-A22B (Jul 2025)",
   "organization": "Alibaba",
   "publication_date": "2025-07-25",
   "parameters": 235000000000.0,
   "training_tokens": 36000000000000.0,
   "tokens_per_param": 153.19148936170214,
   "confidence": "Likely"
  },
  {
   "model": "GLM-4.5",
   "organization": "Z.ai (Zhipu AI),Tsinghua University",
   "publication_date": "2025-07-28",
   "parameters": 355000000000.0,
   "training_tokens": 23100000000000.0,
   "tokens_per_param": 65.07042253521126,
   "confidence": "Confident"
  },
  {
   "model": "LongCat-Flash",
   "organization": "Meituan Inc",
   "publication_date": "2025-09-01",
   "parameters": 560000000000.0,
   "training_tokens": 23000000000000.0,
   "tokens_per_param": 41.07142857142857,
   "confidence": "Likely"
  },
  {
   "model": "Qwen3-Max",
   "organization": "Alibaba",
   "publication_date": "2025-09-05",
   "parameters": 1000000000000.0,
   "training_tokens": 36000000000000.0,
   "tokens_per_param": 36.0,
   "confidence": "Speculative"
  },
  {
   "model": "GLM-4.6",
   "organization": "Z.ai (Zhipu AI),Tsinghua University",
   "publication_date": "2025-09-30",
   "parameters": 357000000000.0,
   "training_tokens": 23000000000000.0,
   "tokens_per_param": 64.42577030812325,
   "confidence": "Likely"
  },
  {
   "model": "Ling-1T",
   "organization": "Ant Group",
   "publication_date": "2025-10-10",
   "parameters": 1000000000000.0,
   "training_tokens": 20000000000000.0,
   "tokens_per_param": 20.0,
   "confidence": "Confident"
  },
  {
   "model": "Olmo 3",
   "organization": "Allen Institute for AI",
   "publication_date": "2025-11-20",
   "parameters": 32000000000.0,
   "training_tokens": 5500000000000.0,
   "tokens_per_param": 171.875,
   "confidence": "Confident"
  },
  {
   "model": "Solar Open 100B",
   "organization": "Upstage",
   "publication_date": "2025-12-31",
   "parameters": 102000000000.0,
   "training_tokens": 19700000000000.0,
   "tokens_per_param": 193.1372549019608,
   "confidence": "Confident"
  },
  {
   "model": "K-EXAONE",
   "organization": "LG AI Research",
   "publication_date": "2025-12-31",
   "parameters": 236000000000.0,
   "training_tokens": 11000000000000.0,
   "tokens_per_param": 46.61016949152543,
   "confidence": "Confident"
  },
  {
   "model": "MiMo-V2.5-Pro",
   "organization": "Xiaomi Corp",
   "publication_date": "2026-04-23",
   "parameters": 1020000000000.0,
   "training_tokens": 27000000000000.0,
   "tokens_per_param": 26.47058823529412,
   "confidence": "Likely"
  },
  {
   "model": "Solar Open2 250B",
   "organization": "Upstage",
   "publication_date": "2026-06-28",
   "parameters": 250300000000.0,
   "training_tokens": 11000000000000.0,
   "tokens_per_param": 43.947263284059126,
   "confidence": "Likely"
  },
  {
   "model": "LongCat-2.0",
   "organization": "Meituan Inc",
   "publication_date": "2026-06-29",
   "parameters": 1600000000000.0,
   "training_tokens": 35000000000000.0,
   "tokens_per_param": 21.875,
   "confidence": "Likely"
  }
 ]
}