{
  "claim": "Jev-Omni ships 11,907,350,320 parameters of fine-tuned text decoder and 983,296 parameters of decision head. Everything that makes it multimodal — the pixel patch embedder, the vision projection, the audio projection — is 52,379,904 parameters of stock Gemma 4 that the release never touched: 0.44% of the model, downloaded from google/gemma-4-12B-it at load time.",
  "method": "Read out of the artifacts, not out of the card. Each of the thirteen backbone shards and the base model's single shard was range-requested for its first eight bytes (the safetensors header length), then for the header itself, and every tensor's shape multiplied out and summed. head.pt is a torch zip archive: its four storage entries are 15,360 / 15,360 / 3,932,160 / 1,024 bytes, which at fp32 is mu[1,3840], sd[1,3840], linear.weight[256,3840] and linear.bias[256] — a Linear(3840 → 256) and two standardisation buffers.",
  "source": "https://huggingface.co/akhilaaa3/Jev-Omni",
  "captured": "2026-09-22",
  "note": "The fine-tuned decoder has exactly the same parameter count as the stock Gemma 4 language model, which is what a merged LoRA should look like: values changed, shapes did not. The backbone is stored in fp32 — 47.6 GB — because the repository's own verification.json records that merging the adapter in lower precision moved a probability by 0.201, against 3.39e-05 in fp32.",
  "columns": [
    { "key": "component", "label": "component", "align": "left" },
    { "key": "params", "label": "parameters", "align": "right", "mono": true },
    { "key": "dtype", "label": "dtype", "align": "right", "mono": true },
    { "key": "origin", "label": "origin", "align": "left" }
  ],
  "rows": [
    { "component": "text decoder, 48 layers (666 tensors, 13 shards)", "params": "11,907,350,320", "dtype": "F32", "origin": "fine-tuned and merged in this repo" },
    { "component": "— of which the token embedding, 248,320 × 3,840", "params": "1,006,632,960", "dtype": "F32", "origin": "fine-tuned and merged in this repo" },
    { "component": "decision head, Linear(3840 → 256)", "params": "983,296", "dtype": "F32", "origin": "trained in this repo (head.pt)" },
    { "component": "vision patch embedder, 48×48 pixels → one token", "params": "35,176,704", "dtype": "BF16", "origin": "stock google/gemma-4-12B-it" },
    { "component": "vision → decoder projection, 3,840 × 3,840", "params": "14,745,600", "dtype": "BF16", "origin": "stock google/gemma-4-12B-it" },
    { "component": "audio → decoder projection, 640 × 3,840", "params": "2,457,600", "dtype": "BF16", "origin": "stock google/gemma-4-12B-it" },
    { "component": "everything the classifier runs, total", "params": "11,960,713,520", "dtype": "—", "origin": "—" }
  ]
}