初始化项目，由ModelHub XC社区提供模型

Model: bineric/lynx-instruct-30b Source: Original Platform
2026-05-02 00:09:18 +08:00
commit fd6872ac04
28 changed files with 171230 additions and 0 deletions
--- a/.gitattributes
+++ b/.gitattributes
@@ -0,0 +1,36 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text
--- a/README.md
+++ b/README.md
@@ -0,0 +1,257 @@
+---
+license: apache-2.0
+language:
+  - en
+  - no
+  - sv
+  - da
+  - is
+base_model: Qwen/Qwen3-30B-A3B-Instruct-2507
+tags:
+  - european
+  - nordic
+  - norwegian
+  - swedish
+  - danish
+  - icelandic
+  - multilingual
+  - moe
+  - qwen3
+library_name: transformers
+pipeline_tag: text-generation
+---
+
+# Bineric Lynx Instruct 30B
+
+**A European large language model with exceptional Nordic language performance.**
+
+| | |
+|---|---|
+| **Parameters** | 30B total, ~3B active |
+| **Architecture** | Qwen3 MoE (128 experts, 8 active) |
+| **Context Length** | 262K tokens |
+| **Base Model** | [Qwen3-30B-A3B-Instruct](https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507) |
+| **Languages** | Norwegian (Bokmål/Nynorsk), Swedish, Danish, Icelandic + 100+ via base |
+
+## About Bineric
+
+[Bineric](https://bineric.com) is an AI company based in Oslo, Norway, built from a European perspective. We started Bineric to make AI usable for organizations that care about governance, language, and where their systems and data actually live.
+
+Lynx is our flagship model — designed to serve European users with strong multilingual support and exceptional Nordic language performance.
+
+## Overview
+
+Lynx is built on Qwen3-30B's efficient Mixture-of-Experts architecture. It retains strong multilingual capabilities across **100+ languages including all major European languages**, while being specifically fine-tuned and rigorously evaluated on Nordic languages (Norwegian, Swedish, Danish, Icelandic) where it demonstrates exceptional results.
+
+**Key features:**
+- Strong European language support inherited from Qwen3 base model
+- Fine-tuned and optimized for Nordic language understanding and generation
+- Efficient MoE architecture: only 3B parameters active per token
+- Available in 8-bit and 4-bit quantized variants for flexible deployment
+- 262K context window for long-document processing
+
+## Try Lynx
+
+Lynx is available through multiple channels:
+
+| Access Method | Link | Best For |
+|--------------|------|----------|
+| **Chatbot** | [chat.bineric.com](https://chat.bineric.com) | Interactive conversations, quick testing |
+| **API** | [bineric.com/platform](https://bineric.com/platform) | Production integrations, programmatic access |
+| **Hugging Face** | This repository | Self-hosting, fine-tuning, research |
+
+## Evaluation Results
+
+Evaluated using [EuroEval](https://github.com/ScandEval/EuroEval) benchmark framework (March 2026).
+
+> **Note:** While Lynx supports all European languages via its Qwen3 base, we have rigorously evaluated performance on Nordic languages. Benchmarks for additional European languages coming soon.
+
+### Nordic Language Performance
+
+| Language | Overall Score | Best Task | Score |
+|----------|---------------|-----------|-------|
+| Danish | **79.3%** | Citizen Tests (Knowledge) | 79.3% |
+| Swedish | **76.9%** | European Values | 76.9% |
+| Norwegian | **71.0%** | NER Nynorsk | 71.0% |
+| Icelandic | **65.1%** | Summarization | 65.1% |
+
+![Language Performance Comparison](assets/language_comparison.svg)
+
+### Task Performance by Language
+
+#### Norwegian (8-bit)
+
+| Task | Dataset | Metric | Score |
+|------|---------|--------|-------|
+| Sentiment | NoReC | MCC | 51.0% |
+| NER (Bokmål) | NorNE-nb | F1 | 65.7% |
+| NER (Nynorsk) | NorNE-nn | F1 | 71.0% |
+| Reading Comprehension | NorQuAD | F1 | 61.2% |
+| Summarization | NoSammendrag | BERTScore | 63.4% |
+| Common Sense | NorCommonSenseQA | MCC | 69.3% |
+| Knowledge | NRK Quiz QA | MCC | 35.3% |
+
+#### Danish (8-bit)
+
+| Task | Dataset | Metric | Score |
+|------|---------|--------|-------|
+| Sentiment | AngryTweets | MCC | 54.8% |
+| NER | DANSK | F1 | 53.8% |
+| Reading Comprehension | MultiWikiQA-da | F1 | **72.2%** |
+| Summarization | Nordjylland News | BERTScore | 65.2% |
+| Common Sense | HellaSwag-da | MCC | 67.7% |
+| Knowledge | Danish Citizen Tests | MCC | **79.3%** |
+| Idioms | Danske Talemåder | MCC | 64.9% |
+
+#### Swedish (8-bit)
+
+| Task | Dataset | Metric | Score |
+|------|---------|--------|-------|
+| Sentiment | SweReC | MCC | 34.5% |
+| NER | SUC3 | F1 | 65.0% |
+| Reading Comprehension | MultiWikiQA-sv | F1 | **72.4%** |
+| Summarization | SweDN | BERTScore | 65.9% |
+| Common Sense | HellaSwag-sv | MCC | 58.3% |
+| Knowledge | MMLU-sv | MCC | 53.9% |
+| European Values | VaLEU-sv | MCC | **76.9%** |
+
+#### Icelandic (8-bit)
+
+| Task | Dataset | Metric | Score |
+|------|---------|--------|-------|
+| NER | MIM-GOLD-NER | F1 | 63.6% |
+| Reading Comprehension | NQiI | F1 | 58.6% |
+| Summarization | RRN | BERTScore | **65.1%** |
+| Knowledge | Icelandic Knowledge | MCC | 28.2% |
+| Common Sense | Winogrande-is | MCC | 9.7% |
+
+![Task Performance by Language](assets/task_performance.svg)
+
+### Quantization Comparison (Norwegian)
+
+8-bit quantization consistently outperforms 4-bit by ~2% on average.
+
+| Task | 4-bit | 8-bit | Delta |
+|------|-------|-------|-------|
+| Sentiment (NoReC) | 49.7% | 51.0% | +1.3% |
+| NER Bokmål | 65.1% | 65.7% | +0.6% |
+| NER Nynorsk | 69.9% | 71.0% | +1.1% |
+| Reading Comp | 58.9% | 61.2% | +2.3% |
+| Summarization | 63.1% | 63.4% | +0.3% |
+| Common Sense | 68.5% | 69.3% | +0.8% |
+| Linguistic Accept. | 29.8% | 36.4% | **+6.6%** |
+
+![8-bit vs 4-bit Quantization](assets/quantization_comparison.svg)
+
+## Strengths & Limitations
+
+### Strengths
+
+- **Named Entity Recognition**: Consistently strong across all languages (63-71% F1)
+- **Reading Comprehension**: Excellent for Danish and Swedish (72%+)
+- **Knowledge Tasks**: Outstanding on Danish Citizen Tests (79.3%)
+- **Summarization**: Stable 63-66% BERTScore across all languages
+
+### Limitations
+
+- **Linguistic Acceptability**: Grammatical judgment tasks are weak (10-36% MCC)
+- **Icelandic Common Sense**: Winogrande-is performance is low (9.7%)
+- **Norwegian Idioms**: Room for improvement (17-19% MCC)
+
+## Quantization Options
+
+| Variant | Size | Quality | Use Case |
+|---------|------|---------|----------|
+| **bfloat16** | ~60GB | Best | Research, high-end GPUs |
+| **8-bit** | ~30GB | ~1-2% loss | Production (A10/L4 GPU) |
+| **4-bit** | ~16GB | ~3-5% loss | Cost-optimized (T4 GPU) |
+
+## Usage
+
+### Basic Inference
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model = AutoModelForCausalLM.from_pretrained(
+    "bineric/lynx-instruct-30b",
+    torch_dtype="auto",
+    device_map="auto"
+)
+tokenizer = AutoTokenizer.from_pretrained("bineric/lynx-instruct-30b")
+
+messages = [
+    {"role": "user", "content": "Hva er hovedstaden i Norge?"}
+]
+
+text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+inputs = tokenizer(text, return_tensors="pt").to(model.device)
+outputs = model.generate(**inputs, max_new_tokens=512)
+print(tokenizer.decode(outputs[0], skip_special_tokens=True))
+```
+
+### With Thinking Mode
+
+Lynx supports extended thinking for complex reasoning tasks:
+
+```python
+messages = [
+    {"role": "user", "content": "Forklar forskjellen mellom bokmål og nynorsk."}
+]
+
+text = tokenizer.apply_chat_template(
+    messages,
+    tokenize=False,
+    add_generation_prompt=True,
+    enable_thinking=True  # Enable reasoning mode
+)
+```
+
+### vLLM Deployment
+
+```bash
+vllm serve bineric/lynx-instruct-30b \
+    --tensor-parallel-size 1 \
+    --max-model-len 32768 \
+    --quantization awq  # For 4-bit
+```
+
+## Model Architecture
+
+```
+Qwen3 MoE Architecture
+├── Total Parameters: 30.5B
+├── Active Parameters: ~3B per token
+├── Hidden Layers: 48
+├── Hidden Size: 2048
+├── Attention Heads: 32
+├── KV Heads: 4 (Grouped Query Attention)
+├── Experts: 128 total
+├── Active Experts: 8 per token
+├── Vocab Size: 151,936
+└── Context Length: 262,144 tokens
+```
+
+## Training
+
+Lynx is fine-tuned from [Qwen3-30B-A3B-Instruct](https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507) with additional training on Nordic language data to improve performance on Norwegian, Swedish, Danish, and Icelandic tasks.
+
+## Citation
+
+```bibtex
+@misc{bineric2026lynx,
+  title={Bineric Lynx: A European Large Language Model},
+  author={Bineric AI},
+  year={2026},
+  publisher={Hugging Face},
+  url={https://huggingface.co/bineric/lynx-instruct-30b}
+}
+```
+
+## Links
+
+- [Base Model: Qwen3-30B-A3B-Instruct](https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507)
+
+---
+
+*Built with care in Oslo by [Bineric](https://bineric.com)*
--- a/added_tokens.json
+++ b/added_tokens.json
@@ -0,0 +1,28 @@
+{
+  "</think>": 151668,
+  "</tool_call>": 151658,
+  "</tool_response>": 151666,
+  "<think>": 151667,
+  "<tool_call>": 151657,
+  "<tool_response>": 151665,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}
--- a/assets/language_comparison.svg
+++ b/assets/language_comparison.svg
@@ -0,0 +1,42 @@
+<svg width="600" height="350" xmlns="http://www.w3.org/2000/svg">
+  <style>
+    .title { font: bold 18px sans-serif; fill: #333; }
+    .label { font: 14px sans-serif; fill: #333; }
+    .value { font: bold 14px sans-serif; fill: #fff; }
+    .axis { font: 12px sans-serif; fill: #666; }
+  </style>
+
+  <!-- Background -->
+  <rect width="600" height="350" fill="#fafafa"/>
+
+  <!-- Title -->
+  <text x="300" y="35" text-anchor="middle" class="title">Overall Performance by Language</text>
+  <text x="300" y="55" text-anchor="middle" class="axis">(EuroEval Benchmark, March 2026)</text>
+
+  <!-- Bars -->
+  <!-- Danish: 79.3% -->
+  <rect x="80" y="90" width="396.5" height="50" fill="#2563eb" rx="4"/>
+  <text x="95" y="120" class="value">Danish</text>
+  <text x="460" y="120" class="label">79.3%</text>
+
+  <!-- Swedish: 76.9% -->
+  <rect x="80" y="155" width="384.5" height="50" fill="#3b82f6" rx="4"/>
+  <text x="95" y="185" class="value">Swedish</text>
+  <text x="448" y="185" class="label">76.9%</text>
+
+  <!-- Norwegian: 71.0% -->
+  <rect x="80" y="220" width="355" height="50" fill="#60a5fa" rx="4"/>
+  <text x="95" y="250" class="value">Norwegian</text>
+  <text x="420" y="250" class="label">71.0%</text>
+
+  <!-- Icelandic: 65.1% -->
+  <rect x="80" y="285" width="325.5" height="50" fill="#93c5fd" rx="4"/>
+  <text x="95" y="315" class="value">Icelandic</text>
+  <text x="390" y="315" class="label">65.1%</text>
+
+  <!-- Scale line -->
+  <line x1="80" y1="345" x2="580" y2="345" stroke="#ccc" stroke-width="1"/>
+  <text x="80" y="342" class="axis">0%</text>
+  <text x="330" y="342" class="axis">50%</text>
+  <text x="560" y="342" class="axis">100%</text>
+</svg>
--- a/assets/quantization_comparison.svg
+++ b/assets/quantization_comparison.svg
@@ -0,0 +1,72 @@
+<svg width="700" height="400" xmlns="http://www.w3.org/2000/svg">
+  <style>
+    .title { font: bold 18px sans-serif; fill: #333; }
+    .label { font: 12px sans-serif; fill: #333; }
+    .value { font: 11px sans-serif; fill: #333; }
+    .axis { font: 11px sans-serif; fill: #666; }
+    .legend { font: 13px sans-serif; fill: #333; }
+  </style>
+
+  <!-- Background -->
+  <rect width="700" height="400" fill="#fafafa"/>
+
+  <!-- Title -->
+  <text x="350" y="30" text-anchor="middle" class="title">8-bit vs 4-bit Quantization (Norwegian)</text>
+
+  <!-- Legend -->
+  <rect x="480" y="50" width="16" height="16" fill="#3b82f6"/>
+  <text x="502" y="63" class="legend">8-bit</text>
+  <rect x="560" y="50" width="16" height="16" fill="#94a3b8"/>
+  <text x="582" y="63" class="legend">4-bit</text>
+
+  <!-- Y-axis labels -->
+  <text x="70" y="100" text-anchor="end" class="label">Sentiment</text>
+  <text x="70" y="145" text-anchor="end" class="label">NER (nb)</text>
+  <text x="70" y="190" text-anchor="end" class="label">NER (nn)</text>
+  <text x="70" y="235" text-anchor="end" class="label">Reading</text>
+  <text x="70" y="280" text-anchor="end" class="label">Summary</text>
+  <text x="70" y="325" text-anchor="end" class="label">Common Sense</text>
+  <text x="70" y="370" text-anchor="end" class="label">Ling. Accept.</text>
+
+  <!-- Bars - Sentiment -->
+  <rect x="80" y="85" width="306" height="18" fill="#3b82f6" rx="2"/>
+  <rect x="80" y="105" width="298.2" height="18" fill="#94a3b8" rx="2"/>
+  <text x="392" y="98" class="value">51.0%</text>
+  <text x="384" y="118" class="value">49.7%</text>
+
+  <!-- Bars - NER nb -->
+  <rect x="80" y="130" width="394.2" height="18" fill="#3b82f6" rx="2"/>
+  <rect x="80" y="150" width="390.6" height="18" fill="#94a3b8" rx="2"/>
+  <text x="480" y="143" class="value">65.7%</text>
+  <text x="476" y="163" class="value">65.1%</text>
+
+  <!-- Bars - NER nn -->
+  <rect x="80" y="175" width="426" height="18" fill="#3b82f6" rx="2"/>
+  <rect x="80" y="195" width="419.4" height="18" fill="#94a3b8" rx="2"/>
+  <text x="512" y="188" class="value">71.0%</text>
+  <text x="505" y="208" class="value">69.9%</text>
+
+  <!-- Bars - Reading -->
+  <rect x="80" y="220" width="367.2" height="18" fill="#3b82f6" rx="2"/>
+  <rect x="80" y="240" width="353.4" height="18" fill="#94a3b8" rx="2"/>
+  <text x="453" y="233" class="value">61.2%</text>
+  <text x="439" y="253" class="value">58.9%</text>
+
+  <!-- Bars - Summary -->
+  <rect x="80" y="265" width="380.4" height="18" fill="#3b82f6" rx="2"/>
+  <rect x="80" y="285" width="378.6" height="18" fill="#94a3b8" rx="2"/>
+  <text x="466" y="278" class="value">63.4%</text>
+  <text x="464" y="298" class="value">63.1%</text>
+
+  <!-- Bars - Common Sense -->
+  <rect x="80" y="310" width="415.8" height="18" fill="#3b82f6" rx="2"/>
+  <rect x="80" y="330" width="411" height="18" fill="#94a3b8" rx="2"/>
+  <text x="501" y="323" class="value">69.3%</text>
+  <text x="497" y="343" class="value">68.5%</text>
+
+  <!-- Bars - Linguistic Acceptability (biggest difference) -->
+  <rect x="80" y="355" width="218.4" height="18" fill="#3b82f6" rx="2"/>
+  <rect x="80" y="375" width="178.8" height="18" fill="#94a3b8" rx="2"/>
+  <text x="304" y="368" class="value">36.4% (+6.6%)</text>
+  <text x="264" y="388" class="value">29.8%</text>
+</svg>
--- a/assets/task_performance.svg
+++ b/assets/task_performance.svg
@@ -0,0 +1,78 @@
+<svg width="750" height="450" xmlns="http://www.w3.org/2000/svg">
+  <style>
+    .title { font: bold 18px sans-serif; fill: #333; }
+    .label { font: 11px sans-serif; fill: #333; }
+    .value { font: 10px sans-serif; fill: #333; }
+    .axis { font: 11px sans-serif; fill: #666; }
+    .legend { font: 12px sans-serif; fill: #333; }
+    .grid { stroke: #e5e7eb; stroke-width: 1; }
+  </style>
+
+  <!-- Background -->
+  <rect width="750" height="450" fill="#fafafa"/>
+
+  <!-- Title -->
+  <text x="375" y="30" text-anchor="middle" class="title">Task Performance Across Nordic Languages</text>
+
+  <!-- Legend -->
+  <rect x="200" y="50" width="14" height="14" fill="#1e40af"/>
+  <text x="218" y="62" class="legend">Danish</text>
+  <rect x="300" y="50" width="14" height="14" fill="#3b82f6"/>
+  <text x="318" y="62" class="legend">Swedish</text>
+  <rect x="400" y="50" width="14" height="14" fill="#60a5fa"/>
+  <text x="418" y="62" class="legend">Norwegian</text>
+  <rect x="520" y="50" width="14" height="14" fill="#93c5fd"/>
+  <text x="538" y="62" class="legend">Icelandic</text>
+
+  <!-- Grid lines -->
+  <line x1="100" y1="100" x2="100" y2="400" class="grid"/>
+  <line x1="100" y1="400" x2="700" y2="400" class="grid"/>
+  <line x1="100" y1="325" x2="700" y2="325" class="grid" stroke-dasharray="4"/>
+  <line x1="100" y1="250" x2="700" y2="250" class="grid" stroke-dasharray="4"/>
+  <line x1="100" y1="175" x2="700" y2="175" class="grid" stroke-dasharray="4"/>
+  <line x1="100" y1="100" x2="700" y2="100" class="grid" stroke-dasharray="4"/>
+
+  <!-- Y-axis labels -->
+  <text x="95" y="405" text-anchor="end" class="axis">0%</text>
+  <text x="95" y="330" text-anchor="end" class="axis">25%</text>
+  <text x="95" y="255" text-anchor="end" class="axis">50%</text>
+  <text x="95" y="180" text-anchor="end" class="axis">75%</text>
+  <text x="95" y="105" text-anchor="end" class="axis">100%</text>
+
+  <!-- X-axis labels -->
+  <text x="160" y="420" text-anchor="middle" class="label">NER</text>
+  <text x="280" y="420" text-anchor="middle" class="label">Reading</text>
+  <text x="400" y="420" text-anchor="middle" class="label">Summary</text>
+  <text x="520" y="420" text-anchor="middle" class="label">Knowledge</text>
+  <text x="640" y="420" text-anchor="middle" class="label">Common Sense</text>
+
+  <!-- NER bars (DA:53.8, SV:65, NO:71, IS:63.6) -->
+  <rect x="125" y="238.4" width="18" height="161.6" fill="#1e40af"/>
+  <rect x="145" y="205" width="18" height="195" fill="#3b82f6"/>
+  <rect x="165" y="186" width="18" height="214" fill="#60a5fa"/>
+  <rect x="185" y="209.2" width="18" height="190.8" fill="#93c5fd"/>
+
+  <!-- Reading Comprehension (DA:72.2, SV:72.4, NO:61.2, IS:58.6) -->
+  <rect x="245" y="183.4" width="18" height="216.6" fill="#1e40af"/>
+  <rect x="265" y="182.8" width="18" height="217.2" fill="#3b82f6"/>
+  <rect x="285" y="216.4" width="18" height="183.6" fill="#60a5fa"/>
+  <rect x="305" y="224.2" width="18" height="175.8" fill="#93c5fd"/>
+
+  <!-- Summarization (DA:65.2, SV:65.9, NO:63.4, IS:65.1) -->
+  <rect x="365" y="204.4" width="18" height="195.6" fill="#1e40af"/>
+  <rect x="385" y="202.3" width="18" height="197.7" fill="#3b82f6"/>
+  <rect x="405" y="209.8" width="18" height="190.2" fill="#60a5fa"/>
+  <rect x="425" y="204.7" width="18" height="195.3" fill="#93c5fd"/>
+
+  <!-- Knowledge (DA:79.3, SV:53.9, NO:35.3, IS:28.2) -->
+  <rect x="485" y="161.9" width="18" height="238.1" fill="#1e40af"/>
+  <rect x="505" y="238.3" width="18" height="161.7" fill="#3b82f6"/>
+  <rect x="525" y="294.1" width="18" height="105.9" fill="#60a5fa"/>
+  <rect x="545" y="315.4" width="18" height="84.6" fill="#93c5fd"/>
+
+  <!-- Common Sense (DA:67.7, SV:58.3, NO:69.3, IS:9.7) -->
+  <rect x="605" y="196.9" width="18" height="203.1" fill="#1e40af"/>
+  <rect x="625" y="225.1" width="18" height="174.9" fill="#3b82f6"/>
+  <rect x="645" y="192.1" width="18" height="207.9" fill="#60a5fa"/>
+  <rect x="665" y="370.9" width="18" height="29.1" fill="#93c5fd"/>
+</svg>
--- a/chat_template.jinja
+++ b/chat_template.jinja
@@ -0,0 +1,89 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
--- a/config.json
+++ b/config.json
@@ -0,0 +1,39 @@
+{
+  "_name_or_path": "Qwen/Qwen3-30B-A3B-Instruct-2507",
+  "architectures": [
+    "Qwen3MoeForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "decoder_sparse_step": 1,
+  "dtype": "bfloat16",
+  "eos_token_id": 151645,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 262144,
+  "max_window_layers": 48,
+  "mlp_only_layers": [],
+  "model_type": "qwen3_moe",
+  "moe_intermediate_size": 768,
+  "norm_topk_prob": true,
+  "num_attention_heads": 32,
+  "num_experts": 128,
+  "num_experts_per_tok": 8,
+  "num_hidden_layers": 48,
+  "num_key_value_heads": 4,
+  "output_router_logits": false,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 10000000,
+  "router_aux_loss_coef": 0.001,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "transformers_version": "4.56.1",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}
--- a/generation_config.json
+++ b/generation_config.json
@@ -0,0 +1,13 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "temperature": 0.6,
+  "top_k": 20,
+  "top_p": 0.95,
+  "transformers_version": "4.56.1"
+}
--- a/merges.txt
+++ b/merges.txt
--- a/model-00001-of-00013.safetensors
+++ b/model-00001-of-00013.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ac2e05480a5ea17e516fce05448ad4c21909f156345caf742cb7f03a4306941
+size 4997184968
--- a/model-00002-of-00013.safetensors
+++ b/model-00002-of-00013.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:439b6d7956efa32189e72ffb484d1341d29f580b438ae1bd454831ae3c00269f
+size 4997741608
--- a/model-00003-of-00013.safetensors
+++ b/model-00003-of-00013.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e382c0120311125d4d28abe767a482470c892a8a540abeff87a641408585f8d0
+size 4997742208
--- a/model-00004-of-00013.safetensors
+++ b/model-00004-of-00013.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:476814fd0fcc326b0885fcc94072ef759177d82b1f5b57aa542a27c9b42683fa
+size 4997743184
--- a/model-00005-of-00013.safetensors
+++ b/model-00005-of-00013.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12dce98a19728160a7648d6e7fd528d42408ba9ea6f45665a74b5bff9c29be4b
+size 4997743184
--- a/model-00006-of-00013.safetensors
+++ b/model-00006-of-00013.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c51659adf51af76ff4fbe2d0eeb534b031ac7eac4e24e599413560c69811098a
+size 4997743184
--- a/model-00007-of-00013.safetensors
+++ b/model-00007-of-00013.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a2ea18bcdec85748f862e8e6c24183add57f51822ab1f688db7e8ff96a7aab0
+size 4997743184
--- a/model-00008-of-00013.safetensors
+++ b/model-00008-of-00013.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72ab7385b376e1fa34f301c6f925f1da3a89b549a8060b3aec407649a1573a6c
+size 4997743184
--- a/model-00009-of-00013.safetensors
+++ b/model-00009-of-00013.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7377351f5fd91b742a68618ff9db39afb56d12178a4616c29e285ea48c02759
+size 4997743184
--- a/model-00010-of-00013.safetensors
+++ b/model-00010-of-00013.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3240f5fbae279b3a002cb2241b819a5515e379804a831d1b244eb9afce6f987
+size 4997743184
--- a/model-00011-of-00013.safetensors
+++ b/model-00011-of-00013.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8045febdb4f375b7a94491621cc60f76658193982a41b2cc525f38202215c09
+size 4997743184
--- a/model-00012-of-00013.safetensors
+++ b/model-00012-of-00013.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:135493f2ed08bedd81e7fb07eea872338f8f68ff6a7dd9d6e49bbe860476ae6c
+size 4997743184
--- a/model-00013-of-00013.safetensors
+++ b/model-00013-of-00013.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c8b6778746d3bea52fe78db8e7709d277e80ea48e1c8fe8940d928e6b7168b4
+size 1094220288
--- a/model.safetensors.index.json
+++ b/model.safetensors.index.json
--- a/special_tokens_map.json
+++ b/special_tokens_map.json
@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}
--- a/tokenizer.json
+++ b/tokenizer.json
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4
+size 11422654
--- a/tokenizer_config.json
+++ b/tokenizer_config.json
@@ -0,0 +1,239 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151667": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151668": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
--- a/vocab.json
+++ b/vocab.json