D

Deepseek Llama3.3 Bllossom 70B

Developed by UNIVA-Bllossom
DeepSeek-llama3.3-Bllossom-70B is built upon the DeepSeek-R1-distill-Llama-70B model, aiming to enhance reasoning performance in Korean environments.
Downloads 1,181
Release Time : 2/12/2025

Model Overview

This model overcomes the limitations of the original base model, which primarily relied on English and Chinese training data, through additional training, specifically optimizing Korean reasoning capabilities. The model's internal thought process is conducted in English, with responses output based on the input language.

Model Features

Multilingual reasoning optimization
Specially optimized for Korean reasoning performance while maintaining English capabilities
Chain-of-thought reasoning
Uses English for internal thought processes and outputs responses based on the input language
Diverse domain data
Includes diverse domain training data in addition to STEM field data

Model Capabilities

Korean text generation
English text generation
Complex reasoning
Multilingual processing

Use Cases

Education
Math problem solving
Solving complex mathematical reasoning problems
Capable of accurately calculating and explaining mathematical problems
Research
Theoretical proofs
Providing multiple proof methods for mathematical theorems
Capable of demonstrating proof processes from different perspectives
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase