Index-1.9B is a series of multilingual LLMs including base, chat, and role-playing variants. The base model was pre-trained on 2.8T primarily Chinese and English tokens. Chat variants use supervised fine-tuning (SFT) and direct preference optimization (DPO) for alignment. The role-playing variant incorporates retrieval-augmented generation (RAG) for fewshots customization. A 32K long-context version is also available.