trend-cybertron/Llama-Primus-Nemotron-70B-Instruct

Llama-Primus-Nemotron-70B-Instruct is a 70 billion parameter instruction-tuned language model developed by trend-cybertron, building upon NVIDIA's Llama-3.1-Nemotron-70B-Instruct. It features a 32768 token context length and is specifically optimized for cybersecurity tasks, achieving an 18.18% improvement in aggregate cybersecurity benchmark scores. The model maintains general-purpose instruction following performance while excelling in specialized cyber threat intelligence and security evaluations.