Create README.md

README.md

@@ -0,0 +1,125 @@
+---
+
+comments: |
+
+  This model card describes the Qwen3.6 IQ2_M 2-bit quantized model repository.
+  
+tags:
+
+- gguf
+
+- iq2-m
+
+- qwen
+
+- conversational
+
+---
+
+
+
+# Qwen3.6 - IQ2_M 2-Bit Quantized Model
+
+
+
+## Overview
+
+This repository contains a **Qwen3.6** large language model series that has been quantized to **IQ2_M (Intelligent Quants) 2-bit precision** using the GGUF format. The model retains high-quality performance while significantly reducing memory footprint through advanced mixed-precision quantization techniques.
+
+
+
+## Model Details
+
+| Property | Value |
+
+|----------|-------|
+
+| Architecture | Qwen35Moe |
+
+| Context Length | 262,144 tokens |
+
+| Quantization Scheme | IQ2_M (2-bit) |
+
+| Format | GGUF (split across 2 files) |
+
+| Total Parameters | ~35B (with MoE routing) |
+
+| File Size | ~8.3 GB + ~3.4 GB |
+
+
+
+## IQ2_M Quantization
+
+The **IQ2_M** quantization scheme is part of the Intelligent Quants (IQ) family developed for efficient model inference:
+
+- **Mixed Precision**: Different weights receive varying bit allocations based on their sensitivity and importance
+- 
+- **Block-wise Quantization**: Optimized scaling factors applied across weight blocks
+- 
+- **2-Bit Compression**: Achieves extreme low-bit precision while preserving critical model capabilities
+- 
+- **Smart Allocation**: Critical parameters preserved in higher precision, less important weights packed into minimal bit formats
+- 
+
+
+## Supported Use Cases
+
+This quantized model is designed for:
+
+- Conversational AI applications
+- 
+- Local inference with llama.cpp, LM Studio, Jan, and similar tools
+- 
+- Memory-efficient deployment scenarios
+- 
+- Practical everyday use cases requiring reduced memory footprint
+- 
+
+
+## Usage Instructions
+
+To load this model locally:
+
+```bash
+
+# Using llama.cpp
+
+llama-server --model Qwen3.6-GeekedOutAi-35B-A3B-BF16-IQ2_M.gguf
+
+
+
+# Using LM Studio or compatible tools
+
+Load the GGUF files from this repository
+
+```
+
+
+
+## Technical Notes
+
+- The model is split into two parts (00001-of-00002 and 00002-of-00002) for efficient storage
+- 
+- IQ2_M quantization maintains conversational capability while achieving significant size reduction
+- 
+- Compatible with llama.cpp and local inference frameworks
+- 
+- Uses imatrix-based calibration for optimal quantization quality
+- 
+
+
+## License & Attribution
+
+This model card documents a community-contributed quantized version of Qwen3.6 optimized for practical deployment scenarios.
+
+
+
+
+
+
+
+
+
+
+
+