{
  "id": "llama-4-scout",
  "name": "Llama 4 Scout",
  "developer": "Meta",
  "exact_model_id": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
  "verified_at": "2026-09-27",
  "evidence_level": "source-verified",
  "access": {
    "label": "Gated / license acceptance",
    "gated": true
  },
  "license": {
    "name": "Llama 4 Community License",
    "commercial_use": "Commercial use is contemplated by the license, subject to its conditions and large-user provisions.",
    "eu_note": "No explicit EU territorial prohibition identified in the reviewed Llama 4 license."
  },
  "architecture": {
    "parameters": "17B active / 109B total",
    "type": "Native multimodal Mixture-of-Experts",
    "context": "10M",
    "modalities": "Text + image → text/code"
  },
  "runtimes": [
    "Meta reference implementation",
    "Transformers"
  ],
  "hardware_note": "Meta states the Llama 4 series needs at least 4 GPUs for full BF16 inference with the reference implementation.",
  "primary_source": "https://github.com/meta-llama/llama-models",
  "third_party_runtime_evidence": [],
  "reference_page": "https://openweightmodels.eu/model-llama-4-scout.html",
  "owm_view": "Llama 4 Scout is a useful case study in the difference between technical openness and licensing simplicity. Its weights are available, its 10M context is exceptional, and it is natively multimodal — but the checkpoint sits under Meta’s custom Llama 4 Community License rather than a standard permissive license. OWM therefore views it as technically portable but legally more conditional than Apache/MIT alternatives.",
  "seo_reference_verified_at": "2026-09-27",
  "change_history": "https://openweightmodels.eu/change-llama-4-scout.json"
}