K

K2

Developed by LLM360
K2 is a large language model with 65 billion parameters, surpassing Llama 2 70B with a 35% computational advantage through a fully transparent training approach.
Downloads 109
Release Time : 4/17/2024

Model Overview

K2 is a large language model jointly developed by MBZUAI, Petuum, and LLM360, featuring a two-phase training process and supporting English text generation tasks.

Model Features

Full transparency
Open-sourcing all training resources, including code, data, model checkpoints, and intermediate results
Efficient training
Surpassing Llama 2 70B with a 35% computational advantage
Two-phase training
Utilizing an optimized two-phase training process
Comprehensive evaluation
Providing systematic evaluation solutions covering general and vertical domains

Model Capabilities

English text generation
Medical knowledge Q&A
Mathematical problem solving
Programming code generation

Use Cases

Knowledge Q&A
General knowledge Q&A
Answering various common-sense questions
Achieved a standardized score of 22.52 in IFEval evaluation
Professional domains
Medical knowledge application
Handling medical-related questions and literature
Achieved a standardized score of 8.55 in medical evaluation
Legal text processing
Analyzing and generating legal-related texts
Trained using the pile-of-law dataset
Technical applications
Programming assistance
Generating and completing code
Trained using the starcoder dataset
Mathematical problem solving
Solving complex mathematical problems
Achieved a standardized score of 2.04 in Math Lvl 5 evaluation
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase