{
  "id": "gemma-3-27b-it",
  "name": "Gemma 3 27B IT",
  "developer": "Google DeepMind",
  "exact_model_id": "google/gemma-3-27b-it",
  "verified_at": "2026-09-27",
  "evidence_level": "source-verified",
  "access": {
    "label": "Gated / terms acceptance on some distribution channels",
    "gated": true
  },
  "license": {
    "name": "Gemma Terms",
    "commercial_use": "Commercial deployment depends on the Gemma Terms and applicable use restrictions; verify exact terms.",
    "eu_note": "No simple Apache/MIT-style conclusion; review Gemma Terms for the deployment context."
  },
  "architecture": {
    "parameters": "27B",
    "type": "Dense multimodal model",
    "context": "128K",
    "modalities": "Image + text → text"
  },
  "runtimes": [
    "Transformers",
    "llama.cpp via GGUF",
    "Ollama"
  ],
  "hardware_note": "Google positions the 27B model as runnable on a single GPU/TPU host; quantization materially changes memory requirements.",
  "primary_source": "https://deepmind.google/models/gemma/gemma-3/",
  "third_party_runtime_evidence": [
    "tp-gemma3-27b-4090"
  ],
  "reference_page": "https://openweightmodels.eu/model-gemma-3-27b-it.html",
  "owm_view": "Gemma 3 27B is interesting because it brings native image understanding, a 128K context window and a relatively workstation-friendly parameter count into the same open-weight package. OWM also considers it a useful reminder that “open weights” and “standard open-source license” are different concepts: Gemma uses its own Terms of Use rather than Apache or MIT.",
  "seo_reference_verified_at": "2026-09-27",
  "change_history": "https://openweightmodels.eu/change-gemma-3-27b-it.json"
}