mudler/carnice-moe-35b-a3b-apex-gguf overview

{
  "metadata": {},
  "card_data": {
    "license": "apache-2.0",
    "base_model": "samuelcardillo/Carnice-MoE-35B-A3B",
    "tags": [
      "gguf",
      "quantized",
      "apex",
      "moe",
      "mixture-of-experts",
      "qwen3.5",
      "agentic",
      "tool-calling"
    ],
    "frontmatter": {
      "license": "apache-2.0",
      "base_model": "samuelcardillo/Carnice-MoE-35B-A3B",
      "tags": [
        "gguf",
        "quantized",
        "apex",
        "moe",
        "mixture-of-experts",
        "qwen3.5",
        "agentic",
        "tool-calling"
      ]
    },
    "hero_image_url": "",
    "summary": "**APEX (Adaptive Precision for EXpert Models)** quantizations of samuelcardillo/Carnice-MoE-35B-A3B. **Brought to you by the LocalAI team** | APEX Project",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nlicense: apache-2.0\nbase_model: samuelcardillo/Carnice-MoE-35B-A3B\ntags:\n  - gguf\n  - quantized\n  - apex\n  - moe\n  - mixture-of-experts\n  - qwen3.5\n  - agentic\n  - tool-calling\n---\n\n# Carnice MoE 35B-A3B APEX GGUF\n\n**APEX (Adaptive Precision for EXpert Models)** quantizations of [samuelcardillo/Carnice-MoE-35B-A3B](https://huggingface.co/samuelcardillo/Carnice-MoE-35B-A3B).\n\n**Brought to you by the [LocalAI](https://github.com/mudler/LocalAI) team** | [APEX Project](https://github.com/mudler/apex-quant)\n\n## Available Files\n\n| File | Profile | Size | Best For |\n|------|---------|------|----------|\n| Carnice-MoE-35B-A3B-APEX-I-Quality.gguf | I-Quality | 21 GB | Highest quality with imatrix |\n| Carnice-MoE-35B-A3B-APEX-Quality.gguf | Quality | 21 GB | Highest quality standard |\n| Carnice-MoE-35B-A3B-APEX-I-Balanced.gguf | I-Balanced | 24 GB | Best overall quality/size ratio |\n| Carnice-MoE-35B-A3B-APEX-Balanced.gguf | Balanced | 24 GB | General purpose |\n| Carnice-MoE-35B-A3B-APEX-I-Compact.gguf | I-Compact | 16 GB | Consumer GPUs, best quality/size |\n| Carnice-MoE-35B-A3B-APEX-Compact.gguf | Compact | 16 GB | Consumer GPUs |\n| Carnice-MoE-35B-A3B-APEX-I-Mini.gguf | I-Mini | 13 GB | Smallest viable, fastest inference |\n| Carnice-MoE-35B-A3B-F16.gguf | F16 | 65 GB | Full precision reference |\n\n## Benchmark Results (Native Evals)\n\n| Model | Size | PPL ↓ | KL ↓ | HellaSwag | WinoGrande | MMLU | ARC-C | TruthfulQA | pp512 t/s | tg128 t/s |\n|:------|-----:|------:|-----:|----------:|-----------:|-----:|------:|-----------:|----------:|----------:|\n| **F16 (ref)** | 65G | 6.16 | - | - | - | - | - | - | 2315 | 109.1 |\n| **APEX-Quality** | 21G | 6.2 | 0.010 | 83.5 | 74.0 | 40.9 | 56.9 | 34.0 | 4717 | 134.2 |\n| **APEX-I-Quality** | 21G | 6.2 | 0.009 | 83.0 | 75.0 | 40.3 | 55.5 | 34.3 | 4734 | 132.6 |\n| **APEX-Balanced** | 24G | 6.2 | 0.007 | 83.0 | 73.8 | 41.1 | 54.5 | 33.8 | 4572 | 130.3 |\n| **APEX-I-Balanced** | 24G | 6.2 | 0.006 | 83.5 | 74.8 | 40.6 | 54.2 | 34.0 | 4539 | 128.7 |\n| **APEX-Compact** | 16G | 6.4 | 0.045 | 82.8 | 75.5 | 40.8 | 55.9 | 34.0 | 4516 | 132.1 |\n| **APEX-I-Compact** | 16G | 6.3 | 0.032 | 83.0 | 73.8 | 41.2 | 56.2 | 34.9 | 4352 | 130.6 |\n| **APEX-I-Mini** | 13G | 6.6 | 0.071 | 82.0 | 72.2 | 40.6 | 53.8 | 33.7 | 4293 | 133.1 |\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.\n\n## Architecture\n\n- **Base Model**: [samuelcardillo/Carnice-MoE-35B-A3B](https://huggingface.co/samuelcardillo/Carnice-MoE-35B-A3B)\n- **Architecture**: Qwen3.5-MoE 35B-A3B\n- **Layers**: 40\n- **Experts**: 256 routed (8 active per token)\n- **Total Parameters**: 35B\n- **Active Parameters**: ~3B per token\n- **APEX Config**: 6+6 symmetric edge gradient across 40 layers\n- **Calibration**: v1.2 diverse dataset\n\n## Run with LocalAI\n\n```bash\nlocal-ai run mudler/Carnice-MoE-35B-A3B-APEX-GGUF@Carnice-MoE-35B-A3B-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",
    "qwen3.5",
    "agentic",
    "tool-calling",
    "base_model:samuelcardillo/Carnice-MoE-35B-A3B",
    "base_model:quantized:samuelcardillo/Carnice-MoE-35B-A3B",
    "license:apache-2.0",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 9,
  "downloads": 4005,
  "gated": false,
  "private": false,
  "last_modified": "2026-04-10T10:19:07.000Z",
  "created_at": "2026-04-10T10:00:51.000Z",
  "pipeline_tag": "",
  "library_name": ""
}

{
  "_id": "69d8ca5334d52e2ae68d53a8",
  "id": "mudler/Carnice-MoE-35B-A3B-APEX-GGUF",
  "modelId": "mudler/Carnice-MoE-35B-A3B-APEX-GGUF",
  "sha": "8bbd0e7ecbdde29175e54cb48ca677a84ba4c515",
  "createdAt": "2026-04-10T10:00:51.000Z",
  "lastModified": "2026-04-10T10:19:07.000Z",
  "author": "mudler",
  "downloads": 4005,
  "likes": 9,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "",
  "siblings_count": 10
}