mudler/nvidia-nemotron-3-super-120b-a12b-apex-gguf overview

{
  "metadata": {},
  "card_data": {
    "license": "other",
    "base_model": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
    "tags": [
      "gguf",
      "quantized",
      "apex",
      "moe",
      "mixture-of-experts",
      "nvidia",
      "nemotron",
      "mamba",
      "hybrid"
    ],
    "frontmatter": {
      "license": "other",
      "base_model": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
      "tags": [
        "gguf",
        "quantized",
        "apex",
        "moe",
        "mixture-of-experts",
        "nvidia",
        "nemotron",
        "mamba",
        "hybrid"
      ]
    },
    "hero_image_url": "",
    "summary": "**APEX (Adaptive Precision for EXpert Models)** quantizations of NVIDIA-Nemotron-3-Super-120B-A12B. **Brought to you by the LocalAI team** | APEX Project | Technical Report",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nlicense: other\nbase_model: nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16\ntags:\n  - gguf\n  - quantized\n  - apex\n  - moe\n  - mixture-of-experts\n  - nvidia\n  - nemotron\n  - mamba\n  - hybrid\n---\n\n# Nemotron-3-Super-120B-A12B APEX GGUF\n\n**APEX (Adaptive Precision for EXpert Models)** quantizations of [NVIDIA-Nemotron-3-Super-120B-A12B](https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16).\n\n**Brought to you by the [LocalAI](https://github.com/mudler/LocalAI) team** | [APEX Project](https://github.com/mudler/apex-quant) | [Technical Report](https://github.com/mudler/apex-quant/blob/main/paper/APEX_Technical_Report.pdf)\n\n## Benchmark Results\n\nBenchmarks coming soon. For reference APEX benchmarks on the Qwen3.5-35B-A3B architecture, see [mudler/Qwen3.5-35B-A3B-APEX-GGUF](https://huggingface.co/mudler/Qwen3.5-35B-A3B-APEX-GGUF).\n\n## What is APEX?\n\nAPEX is a quantization strategy for Mixture-of-Experts (MoE) models. It classifies tensors by role (routed expert, shared expert, attention) and applies a layer-wise precision gradient -- edge layers get higher precision, middle layers get more aggressive compression. I-variants use diverse imatrix calibration (chat, code, reasoning, tool-calling, agentic traces, Wikipedia).\n\nSee the [APEX project](https://github.com/mudler/apex-quant) for full details, technical report, and scripts.\n\n## Architecture\n\n- **Model**: NVIDIA-Nemotron-3-Super-120B-A12B (NemotronH)\n- **Layers**: 88\n- **Type**: Hybrid Mamba-2 / LatentMoE / Attention + Multi-Token Prediction (MTP)\n- **Experts**: 512 routed + 1 shared (22 active per token)\n- **Total Parameters**: 120B\n- **Active Parameters**: ~12B per token\n- **APEX Config**: 5+5 symmetric edge gradient across 88 layers\n\n## Run with LocalAI\n\n```bash\nlocal-ai run mudler/NVIDIA-Nemotron-3-Super-120B-A12B-APEX-GGUF@Nemotron-3-Super-120B-A12B-APEX-I-Balanced.gguf\n```\n\n## Credits\n\nAPEX is brought to you by the [LocalAI](https://github.com/mudler/LocalAI) team. Developed through human-driven, AI-assisted research. Built on [llama.cpp](https://github.com/ggerganov/llama.cpp).\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "quantized",
    "apex",
    "moe",
    "mixture-of-experts",
    "nvidia",
    "nemotron",
    "mamba",
    "hybrid",
    "base_model:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
    "base_model:quantized:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
    "license:other",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 1,
  "downloads": 1985,
  "gated": false,
  "private": false,
  "last_modified": "2026-04-05T23:39:28.000Z",
  "created_at": "2026-04-05T23:06:23.000Z",
  "pipeline_tag": "",
  "library_name": ""
}

{
  "_id": "69d2eaef787b38cf81d8d94b",
  "id": "mudler/NVIDIA-Nemotron-3-Super-120B-A12B-APEX-GGUF",
  "modelId": "mudler/NVIDIA-Nemotron-3-Super-120B-A12B-APEX-GGUF",
  "sha": "ee73f6cf76040e0313c054031ad338f43741834a",
  "createdAt": "2026-04-05T23:06:23.000Z",
  "lastModified": "2026-04-05T23:39:28.000Z",
  "author": "mudler",
  "downloads": 1985,
  "likes": 1,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "",
  "siblings_count": 9
}