{
  "modelName": "functiongemma-270m-it",
  "modelFamily": "functiongemma",
  "format": "coreml-mlprogram",
  "architecture": "fixed_context_last_logits",
  "source": "huggingface",
  "hfRepo": "google/functiongemma-270m-it",
  "maxSeqLen": 128,
  "computePrecision": "float32",
  "quantization": "int8",
  "inputs": [
    "input_ids",
    "causal_mask",
    "cos_full",
    "sin_full",
    "cos_sliding",
    "sin_sliding"
  ],
  "output": "logits",
  "limitations": [
    "full-context recompute per generated token",
    "no KV cache",
    "prompts longer than maxSeqLen are truncated from the left"
  ]
}