{
 "slug": "GPT-1",
 "title": "GPT-1",
 "type": "model",
 "short_desc": "OpenAI's June 2018 Generative Pre-trained Transformer, the 117-million-parameter model that established the pretrain-then-finetune recipe behind the GPT series.",
 "categories": [
  "OpenAI models",
  "Models",
  "2018 model releases",
  "Open-weight models"
 ],
 "infobox": {
  "Developer": "OpenAI",
  "Family": "GPT",
  "Announced": "June 2018",
  "Architecture": "12-layer decoder-only Transformer",
  "Parameters": "117 million",
  "Context window": "512 tokens",
  "Post-training": "Supervised fine-tuning on individual downstream tasks",
  "License / access": "Code and weights released publicly (MIT license)",
  "Paper/report": "Radford et al., \"Improving Language Understanding by Generative Pre-Training\" (June 2018)"
 },
 "infobox_links": {
  "Developer": [
   "OpenAI"
  ],
  "Family": [
   "GPT_(model_family)"
  ],
  "Architecture": [
   "Transformer_(architecture)"
  ],
  "Post-training": [
   "Instruction_tuning"
  ]
 },
 "aliases": [
  "GPT"
 ],
 "family": "GPT (model family)",
 "org": "OpenAI",
 "registry_status": "stub",
 "words": 400,
 "references": 3,
 "outbound": [
  "Anthropic",
  "BERT",
  "GPT-2",
  "GPT-3",
  "GPT-3.5",
  "GPT-4",
  "GPT_(model_family)",
  "Instruction_tuning",
  "Large_language_model",
  "OpenAI",
  "Pretraining",
  "Scaling_laws",
  "Self-attention",
  "Supervised_fine-tuning",
  "Tokenization",
  "Transformer_(architecture)"
 ],
 "inbound": [
  "DistilBERT",
  "ELECTRA",
  "ELMo",
  "Feed-forward_network",
  "GPT-2",
  "GPT_(model_family)",
  "GloVe",
  "Layer_normalization",
  "Masked_language_modeling",
  "Positional_encoding",
  "Pretraining",
  "Removal_of_Sam_Altman_from_OpenAI_(annotated_edition)",
  "RoBERTa",
  "Supervised_fine-tuning",
  "Transformer_(architecture)",
  "ULMFiT"
 ],
 "url": "wiki/GPT-1.html",
 "built": "2026-07-24"
}
