# Multi-task instruction following

A large language model for network security continuously trained based on nvidia/Llama-3.1-Nemotron-70B-Instruct, with a 18.18% increase in the comprehensive score in network security benchmark tests.

Tülu 3 is a leading family of instruction-following models, offering fully open-source data, code, and training methodologies as a comprehensive guide to modern technology. Version 3.1 features improvements in the reinforcement learning phase, delivering enhanced overall performance.

An instruction-following model fine-tuned based on OpenLlama-3B, optimized for role-playing, instruction following, and code generation