{
  "id": "opt-6-7b",
  "name": "OPT 6.7B",
  "identifiers": {
    "huggingface": [
      "facebook/opt-6.7b"
    ]
  },
  "developer": "Meta AI",
  "release_date": {
    "value": "2022-05-02",
    "source": "https://arxiv.org/abs/2205.01068",
    "status": "partial",
    "note": "arXiv v1 date of the paper."
  },
  "weights_status": "open",
  "availability": {
    "value": "available",
    "checked": "2026-09-24",
    "source": "https://huggingface.co/facebook/opt-6.7b"
  },
  "license": {
    "value": "OPT-175B License Agreement: non-commercial research use only.",
    "source": "https://github.com/facebookresearch/metaseq/blob/main/projects/OPT/MODEL_LICENSE.md",
    "status": "recorded",
    "note": "metaseq OPT README: 'The use of OPT model weights is subject to the Model License.'"
  },
  "architecture": {
    "family": "decoder_only",
    "note": "Paper: trained 'to roughly match the performance and sizes of the GPT-3 class of models'; hyperparameters 'largely follow Brown et al. (2020)'. ReLU activation. The HF model card for this repo was written by Hugging Face, not Meta; Meta's own card is Appendix D of the paper.",
    "n_layers": {
      "value": 32,
      "source": "https://huggingface.co/facebook/opt-6.7b/raw/a45aa65bbeb77c1558bc99bedc6779195462dab0/config.json",
      "status": "recorded"
    },
    "hidden_size": {
      "value": 4096,
      "source": "https://huggingface.co/facebook/opt-6.7b/raw/a45aa65bbeb77c1558bc99bedc6779195462dab0/config.json",
      "status": "recorded"
    },
    "n_heads": {
      "value": 32,
      "source": "https://huggingface.co/facebook/opt-6.7b/raw/a45aa65bbeb77c1558bc99bedc6779195462dab0/config.json",
      "status": "recorded"
    },
    "vocab_size": {
      "value": 50272,
      "source": "https://huggingface.co/facebook/opt-6.7b/raw/a45aa65bbeb77c1558bc99bedc6779195462dab0/config.json",
      "status": "recorded",
      "note": "Paper: GPT-2 byte-level BPE tokenizer."
    },
    "positional_encoding": {
      "value": "learned_absolute",
      "source": "https://github.com/huggingface/transformers/blob/v4.34.0/src/transformers/models/opt/modeling_opt.py",
      "status": "recorded",
      "note": "Code: OPTLearnedPositionalEmbedding."
    },
    "context_length": {
      "value": 2048,
      "source": "https://arxiv.org/abs/2205.01068",
      "status": "recorded"
    }
  },
  "training_data": {
    "value": "~180B tokens: RoBERTa corpus subsets (BookCorpus, Stories, updated CCNews), a subset of the Pile, and PushShift.io Reddit; tokenized with the GPT-2 BPE.",
    "source": "https://arxiv.org/abs/2205.01068",
    "status": "recorded"
  },
  "techniques": [
    "transformer-decoder",
    "byte-pair-encoding"
  ],
  "primary_sources": [
    "https://arxiv.org/abs/2205.01068",
    "https://github.com/facebookresearch/metaseq/blob/main/projects/OPT/README.md",
    "https://github.com/facebookresearch/metaseq/blob/main/projects/OPT/MODEL_LICENSE.md",
    "https://huggingface.co/facebook/opt-6.7b/raw/a45aa65bbeb77c1558bc99bedc6779195462dab0/config.json"
  ],
  "record_history": [
    {
      "date": "2026-09-24",
      "change": "created from primary sources (Phase 1 seed, batch 2)",
      "by": "wilson-pruitt + claude"
    },
    {
      "date": "2026-09-24",
      "change": "availability checked and recorded",
      "by": "wilson-pruitt + claude"
    }
  ]
}