QuixiAI/Ina-v11.1

Ina-v11.1 is a 70 billion parameter LLaMA-compatible model developed by BaiAI and Eric Hartford (QuixiAI), fine-tuned for high-fidelity role-playing. It interprets persona definitions as executable instructions, ensuring extreme obedience to character descriptions and forbidden phrases. This model excels in creative and NSFW role-play, maintaining tone and speech patterns over thousands of tokens.