KISTI-KONI/KONI-Llama3.1-8B-R-20250831

TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:Aug 12, 2025License:llama3.1Architecture:Transformer0.0K Cold

KONI-Llama3.1-8B-R-20250831 is an 8 billion parameter instruction-tuned language model developed by KISTI, OneLineAI, HAE-RAE, and Oracle, built upon KONI-Llama3.1-8B-Instruct-20241024. It is optimized for Korean-centric reasoning with full English support, leveraging a Language-Mixed Chain-of-Thought strategy for improved consistency and accuracy. This model excels in scientific and technological domains, general knowledge, mathematics, and logical problem-solving, with a context length of 32768 tokens.

Loading preview...

Overview

KONI-Llama3.1-8B-R-20250831 is an 8 billion parameter model from the KISTI-KONI series, developed collaboratively by KISTI, OneLineAI, HAE-RAE, and Oracle. It is an instruction-tuned variant based on KONI-Llama3.1-8B-Instruct-20241024, specifically designed for Korean-centric reasoning while maintaining full English support. The model utilizes a unique Language-Mixed Chain-of-Thought strategy, which interleaves Korean and English during the reasoning process to enhance consistency and accuracy in complex problem-solving.

Key Capabilities

  • Bilingual Reasoning: Optimized for Korean reasoning tasks with robust English support.
  • Scientific & Technological Specialization: Trained with a strong emphasis on scientific and technological domains, making it suitable for expert-level queries.
  • Enhanced Reasoning Accuracy: Employs a Language-Mixed Chain-of-Thought strategy for improved performance in complex reasoning tasks.
  • Broad Task Coverage: Addresses a wide range of tasks including science and technology queries, general knowledge, mathematics, and logical problem-solving.
  • High Context Length: Supports a context length of 32768 tokens.

Good For

  • Science and technology Q&A in both Korean and English.
  • Mathematical and logical problem-solving requiring complex reasoning.
  • Korean document understanding and analysis.
  • Serving as a reasoning backbone for agent systems.