{
  "id": "llama-2-7b",
  "name": "Llama 2 7B",
  "identifiers": {
    "huggingface": [
      "meta-llama/Llama-2-7b-hf",
      "meta-llama/Llama-2-7b"
    ]
  },
  "developer": "Meta",
  "release_date": {
    "value": "2023-07-18",
    "source": "https://ai.meta.com/blog/llama-2/",
    "status": "recorded"
  },
  "weights_status": "open",
  "availability": {
    "value": "gated",
    "checked": "2026-09-24",
    "source": "https://huggingface.co/meta-llama/Llama-2-7b-hf",
    "note": "HF gate: manual approval."
  },
  "license": {
    "value": "Llama 2 Community License Agreement (research and commercial use; licensees with >700M monthly active users on the release date must request a separate license from Meta).",
    "source": "https://github.com/meta-llama/llama/blob/main/LICENSE",
    "status": "recorded"
  },
  "architecture": {
    "family": "decoder_only",
    "note": "Paper (sec. 2.2): 'We adopt most of the pretraining setting and model architecture from Llama 1'; primary differences are longer context and GQA, and GQA is used only at 34B/70B (Table 1 marks 7B without GQA). Per-size dimensions are not given in the paper, and Meta's HF config is gated, so layers/hidden/heads stay not_recorded rather than copied from Llama 1.",
    "n_layers": {
      "value": null,
      "status": "not_recorded"
    },
    "hidden_size": {
      "value": null,
      "status": "not_recorded"
    },
    "n_heads": {
      "value": null,
      "status": "not_recorded"
    },
    "vocab_size": {
      "value": 32000,
      "source": "https://arxiv.org/abs/2307.09288",
      "status": "recorded",
      "note": "Paper says 'The total vocabulary size is 32k tokens' and 'the same tokenizer as Llama 1'."
    },
    "positional_encoding": {
      "value": "rotary (RoPE)",
      "source": "https://arxiv.org/abs/2307.09288",
      "status": "recorded"
    },
    "context_length": {
      "value": 4096,
      "source": "https://arxiv.org/abs/2307.09288",
      "status": "recorded",
      "note": "Table 1: '4k'."
    }
  },
  "training_data": {
    "value": "2.0T tokens, 'a new mix of data from publicly available sources, which does not include data from Meta's products or services.' Per-source breakdown not disclosed (unlike Llama 1).",
    "source": "https://arxiv.org/abs/2307.09288",
    "status": "partial"
  },
  "techniques": [
    "transformer-decoder",
    "rotary-position-embedding",
    "rmsnorm",
    "swiglu"
  ],
  "primary_sources": [
    "https://arxiv.org/abs/2307.09288",
    "https://ai.meta.com/blog/llama-2/",
    "https://github.com/meta-llama/llama/blob/main/LICENSE"
  ],
  "record_history": [
    {
      "date": "2026-09-24",
      "change": "created from primary sources (Phase 1 seed, batch 1)",
      "by": "wilson-pruitt + claude"
    },
    {
      "date": "2026-09-24",
      "change": "availability checked and recorded",
      "by": "wilson-pruitt + claude"
    }
  ]
}