teamsparta-inc/K-AX-Spartan-OSS-27B
K-AX Spartan OSS 27B is a 27 billion parameter Korean foundation model developed by 스파르타AX (Sparta AX), built on the Qwen 3.5 architecture. It features a 262,144 token context window and is pre-trained on a large-scale multilingual corpus with domain adaptation for Korean-specific knowledge. The model is optimized for Korean unique domains such as legal, professional exams, and culture, and is designed for native compatibility with the K-AX Spartan Adapter product family.
Loading preview...
K-AX Spartan OSS 27B: A Korean Foundation Model
K-AX Spartan OSS 27B is a 27 billion parameter Korean foundation model developed by 스파르타AX (Sparta AX), featuring a substantial 262,144 token context window. Built upon the Qwen 3.5 architecture, this model is the open-weight release within the K-AX Spartan series, designed to serve as the base for the K-AX Spartan Adapter product family.
Key Capabilities & Features
- Korean Domain Optimization: Specialized for Korean-specific knowledge, including legal, professional examinations, and cultural contexts.
- Advanced Training Pipeline: Utilizes continual pre-training on a proprietary multilingual corpus, followed by domain-adaptive pre-training for Korean content, and a multi-stage alignment process.
- Adapter-Native Architecture: Engineered for native compatibility with K-AX Spartan Adapters, allowing for cost-effective domain-specific specialization (e.g., Legal, Medical, Finance).
- Broad Inference Engine Compatibility: Supports major inference engines like vLLM, SGLang, and TensorRT-LLM.
When to Use This Model
- Korean Language Applications: Ideal for tasks requiring deep understanding and generation of Korean text, especially in specialized domains.
- Domain-Specific Customization: Suitable for developers looking to build highly specialized applications using the K-AX Spartan Adapter framework.
- Research and Development: A robust foundation model for academic research, personal learning, and benchmarking in Korean NLP.