Blackroot/Mirai-70B-1.0

Blackroot/Mirai-70B-1.0 is a 70 billion parameter stockmerge model, built upon a Llama-3.1-Nemotron-70B-Instruct base and enhanced with multiple Llama-3.1 and Llama-3 variants. This model excels in generating high-quality, varied prose and intricate narratives, making it particularly suited for complex roleplay and creative writing applications. It demonstrates improved logic, reasoning, and reduced repetition compared to previous iterations, with a 32K token context length.