SKT-NRS/NRS_QWEN_MYTHOS_1M
SKT-NRS/NRS_QWEN_MYTHOS_1M is a 9 billion parameter language model developed by SKT AI Labs, based on the Qwen 3.5 architecture. It features a massive 1 million token context window, achieved through YaRN scaling, and is enhanced with a Neural Reasoning System (NRS) for significantly improved logical thinking and coherence. This model is optimized for deep reasoning tasks, fast inference on consumer hardware, and supports native tool calling for Python execution and web search.
Loading preview...
SKT-NRS/NRS_QWEN_MYTHOS_1M: Enhanced Reasoning with 1M Context
SKT-NRS/NRS_QWEN_MYTHOS_1M is a custom fine-tuned model from SKT AI Labs, built upon the Qwen 3.5 9B architecture. It integrates a proprietary Neural Reasoning System (NRS) to deliver exceptional reasoning depth and coherence, alongside blazing fast inference speeds.
Key Capabilities
- 100x High Reasoning Capacity: Achieved through NRS Boosting, dramatically improving logical thinking.
- 1 Million Token Context: Utilizes YaRN RoPE Scaling to support extensive context windows, ideal for large documents and codebases.
- 10x Thinking Enhancement: Incorporates advanced step-by-step
<think>tags, refined via SKT's Supervised Fine-Tuning (SFT). - Lightning Fast Inference: Optimized for efficient performance on consumer-grade hardware like RTX 3090/4090 GPUs.
- Native Tool Calling: Equipped with enhanced function calling capabilities for Python execution and Web Search.
Training & Optimization
The model underwent a rigorous NRS enhancement pipeline, including the generation of high-quality Chain-of-Thought (CoT) data and fine-tuning on approximately 500k reasoning samples covering coding, mathematics, and logic. It is designed for open research, offering high performance in an efficient 9B parameter package.