fieldmouse-ai/qwen3.5-sovereign-titan-9b-gguf IQ3_XXS GGUF - Free GGUF Download is indexed on GraySoft with repository links, GGUF quant files, and Hugging Face metadata. This page helps you pick a local model for guIDE or other runtimes. See related models in the same shard below.
fieldmouse-ai/qwen3.5-sovereign-titan-9b-gguf overview
> TEXT-ONLY OPTIMIZATION / テキスト専用最適化 This Ollama release is a high-density text-only build. Vision support is excluded to prioritize logical throughput and weight efficiency in the fieldmouse-ai namespace. 👁️ FOR VISION SUPPORT (Multimodal): Please visit our HuggingFace Forge to download the full GGUF weights and accompanying .mmproj files. 🌐 Official Website: FieldMouse-AI.com The Sovereign Series is a precision-engineered distribution of Qwen 3.5, meticulously forged to provide architectural consistency and high-fidelity logic across the entire computational spectrum. Sovereign(主権者)シリーズは、Qwen 3.5を精密に再構築したディストリビューションです。計算資源の全スペクトルにおいて、アーキテクチャの一貫性と高精度な論理を提供するために丹念に鍛造されました。
Repository Files & Downloads
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| MMPROJ-QWEN-QWEN3.5-9B-F16.gguf | GGUF | F16 | 875.63 MB | Download |
| QWEN-QWEN3.5-9B-BF16.gguf | GGUF | BF16 | 16.69 GB | Download |
| QWEN-QWEN3.5-9B-IQ3_M.gguf | GGUF | IQ3_M | 4.11 GB | Download |
| QWEN-QWEN3.5-9B-IQ3_S.gguf | GGUF | IQ3_S | 4.07 GB | Download |
| QWEN-QWEN3.5-9B-IQ3_XS.gguf | GGUF | IQ3_XS | 3.95 GB | Download |
| QWEN-QWEN3.5-9B-IQ3_XXS.gguf | GGUF | IQ3_XXS | 3.67 GB | Download |
| QWEN-QWEN3.5-9B-IQ4_XS.gguf | GGUF | IQ4_XS | 4.84 GB | Download |
| QWEN-QWEN3.5-9B-Q3_K_M.gguf | GGUF | Q3_K_M | 4.31 GB | Download |
| QWEN-QWEN3.5-9B-Q3_K_S.gguf | GGUF | Q3_K_S | 3.97 GB | Download |
| QWEN-QWEN3.5-9B-Q4_K_M.gguf | GGUF | Q4_K_M | 5.24 GB | Download |
| QWEN-QWEN3.5-9B-Q4_K_S.gguf | GGUF | Q4_K_S | 4.98 GB | Download |
| QWEN-QWEN3.5-9B-Q5_K_M.gguf | GGUF | Q5_K_M | 6.02 GB | Download |
| QWEN-QWEN3.5-9B-Q5_K_S.gguf | GGUF | Q5_K_S | 5.87 GB | Download |
| QWEN-QWEN3.5-9B-Q6_K.gguf | GGUF | Q6_K | 6.85 GB | Download |
| QWEN-QWEN3.5-9B-Q8_0.gguf | GGUF | — | 8.87 GB | Download |
Model Details Live
Metadata Inspector
Normalized metadata (stored in metadata_json)
{
"metadata": {},
"card_data": {
"frontmatter": {},
"hero_image_url": "",
"summary": "> [!IMPORTANT] > **TEXT-ONLY OPTIMIZATION / テキスト専用最適化** > > This Ollama release is a high-density text-only build. Vision support is excluded to prioritize logical throughput and weight efficiency in the fieldmouse-ai namespace. > > **👁️ FOR VISION SUPPORT (Multimodal):** > Please visit our HuggingFace Forge to download the full GGUF weights and accompanying .mmproj files. **🌐 Official Website: FieldMouse-AI.com** The **Sovereign Series** is a precision-engineered distribution of Qwen 3.5, meticulously forged to provide architectural consistency and high-fidelity logic across the entire computational spectrum. **Sovereign(主権者)シリーズ**は、Qwen 3.5を精密に再構築したディストリビューションです。計算資源の全スペクトルにおいて、アーキテクチャの一貫性と高精度な論理を提供するために丹念に鍛造されました。",
"quick_links": [],
"benchmark_table_html": "",
"readme_markdown": "---\n{}\n---\n# Qwen 3.5 Sovereign Series: 75-Model Foundry Manifest\n# Qwen 3.5 Sovereignシリーズ: 75モデル・ファウンドリ・マニフェスト\n\n> [!IMPORTANT]\n> **TEXT-ONLY OPTIMIZATION / テキスト専用最適化**\n> \n> This Ollama release is a high-density text-only build. Vision support is excluded to prioritize logical throughput and weight efficiency in the `fieldmouse-ai` namespace.\n> \n> **👁️ FOR VISION SUPPORT (Multimodal):**\n> Please visit our [HuggingFace Forge](https://huggingface.co/FieldMouse-AI) to download the full GGUF weights and accompanying `.mmproj` files.\n\n**🌐 Official Website: [FieldMouse-AI.com](https://FieldMouse-AI.com)**\n\nThe **Sovereign Series** is a precision-engineered distribution of Qwen 3.5, meticulously forged to provide architectural consistency and high-fidelity logic across the entire computational spectrum. \n\n**Sovereign(主権者)シリーズ**は、Qwen 3.5を精密に再構築したディストリビューションです。計算資源の全スペクトルにおいて、アーキテクチャの一貫性と高精度な論理を提供するために丹念に鍛造されました。\n\n\n## 💻 \"High-Fidelity AI for Every Machine\" — 誰もがAIを所有できる世界へ\n\n- **The Hardware (ハードウェア):** \n 私たちの手元にある **「古い鉄(Iron)」**を再定義します。2011年製のMac Mini (6+ t/s) または Ryzen 7 (22+ t/s) から最新のDual RTX 3060 (160+ t/s) まで、あらゆる環境で最高のパフォーマンスを引き出します。 \n Optimized for the 'Iron' we already own. Reclaiming legacy hardware.\n\n- **The Tech (テクノロジー):** \n すべてのモデルは I-Matrix によって精緻にキャリブレーション(校正)されています。これは単なる「軽量化」ではなく、推論の密度(Reasoning Density)を維持するためのプロセスです。 \n These aren't just 'small' models; they are 'dense' models.\n\n- **The Philosophy (フィロソフィー):** \n FieldMouse-AIの使命は、高度なバイリンガル推論(日・英)と、低リソース環境へのアクセシビリティの架け橋となることです。 \n Bridging the gap between high-level bilingual reasoning and low-resource accessibility.\n\n\n\n## ⚗️ The Foundry Methodology / 鍛造のメソッド\n\nStability was achieved through a rigorous, three-stage optimization process:\n厳格な3段階の最適化プロセスにより、圧倒的な安定性を実現しました:\n\n1. **Custom Imatrix Calibration**: Importance matrices were generated using targeted datasets to preserve the \"logical spine\" of the model during quantization.\n **カスタムImatrixキャリブレーション**: ターゲットデータセットを用いた重要度行列(imatrix)の生成により、量子化プロセスにおけるモデルの「論理的骨格」を維持しています。\n2. **Tiered Penalty Scaling**: Unique configurations for `presence_penalty` and `repeat_penalty` were applied to each parameter class to prevent cognitive drift.\n **階層型ペナルティ・スケーリング**: 各パラメータクラスに応じた`presence_penalty`と`repeat_penalty`の独自設定により、認知的ドリフトを防止しています。\n3. **Japanese Language Prioritization**: The `imatrix` process was specifically optimized for high-fidelity Japanese performance, ensuring cultural and linguistic nuance is preserved even at low bitrates.\n **日本語優先最適化**: 日本語のパフォーマンスを最優先にimatrixプロセスを最適化。低ビットレートにおいても、文化的・言語的なニュアンスを高度に維持します。\n\n## 📊 The Fleet Tiers / 艦隊ティア\n\n| Designation / 名称 | Parameters | Focus / 焦点 | Benchmark (TPS) | Size (Avg) / サイズ |\n| :--- | :--- | :--- | :--- | :--- |\n| **GHOST** | 0.8B | Legacy Reclamation | 6+ (CPU) / 130+ (GPU) | < 1 GB |\n| **VANGUARD** | 2B | High-Density Logic | 115+ (GPU) | 1.3 GB |\n| **GUARDIAN** | 4B | Technical Accuracy | 60+ (GPU) | 2.7 GB |\n| **TITAN** | 9B | Deep Architecture | 40+ (GPU) | 5.6 GB |\n| **COLOSSUS** | 27B | Frontier-class Intelligence | 15+ (GPU) | 16.0 GB |\n\n### Key Milestone: The Vanguard Efficiency / ヴァンガードの効率性\n\nThe **Vanguard-2B (IQ3_M)** achieves **124 TPS** while maintaining full C++ systems architecture capabilities and bilingual (EN/JP) fluidity.\n\n**Vanguard-2B (IQ3_M)**は、C++システムアーキテクチャの完全な記述能力と日英バイリンガルの流暢性を維持しつつ、**124 TPS**という驚異的な速度を達成しました。\n\n\n## ⚙️ Configuration Parameters / 設定パラメータ\n\n- **Repeat Penalty**: `1.4` (Ghost/Vanguard), `1.15` (Guardian/Titan/Colossus), \n- **Presence Penalty**: `0.50` (Ghost/Vanguard), `0.10` (Guardian), `0.00` (Titan/Colossus)\n- **Context Length**: `4096`\n\n## 🕊️ The Setomono Spirit / 瀬戸物の精神\n\nThis work is provided as a testament to the reclamation of local silicon. It requires no cloud, no subscription, and no external validation. The iron is yours. The intelligence is Sovereign.\n\nこの成果は、ローカルな「鉄(シリコン)」の奪還の証です。クラウドも、サブスクリプションも、外部の承認も必要ありません。鉄はあなたのもの。知性は、あなたの主権(Sovereign)の下に。\n\n---\n*Forged by the FieldMouse-AI Foundry / FieldMouse-AI鍛造所にて製作*",
"related_quantizations": []
},
"tags": [
"gguf",
"region:us"
],
"likes": 0,
"downloads": 5825,
"gated": false,
"private": false,
"last_modified": "2026-03-30T11:22:59.000Z",
"created_at": "2026-03-19T16:08:44.000Z",
"pipeline_tag": "",
"library_name": ""
}
Source payload excerpt (from Hugging Face API)
{
"_id": "69bc1f8c3ea5f02a29e292dc",
"id": "FieldMouse-AI/Qwen3.5-Sovereign-Titan-9B-GGUF",
"modelId": "FieldMouse-AI/Qwen3.5-Sovereign-Titan-9B-GGUF",
"sha": "ba94f550057bd7cbf486ea29d4302146374944fc",
"createdAt": "2026-03-19T16:08:44.000Z",
"lastModified": "2026-03-30T11:22:59.000Z",
"author": "FieldMouse-AI",
"downloads": 5825,
"likes": 0,
"gated": false,
"private": false,
"pipeline_tag": "",
"library_name": "",
"siblings_count": 17
}