Yi 34B

About

Developed by the Chinese startup 01.AI, the Yi 34B model is a robust large language model known for its bilingual capabilities in English and Chinese. It leverages a standard Transformer architecture, distinctively trained from scratch, featuring innovations like Grouped-Query Attention (GQA), SwiGLU activation, and RoPE with an adjusted base frequency. Yi 34B excels in diverse natural language processing tasks, such as language understanding and commonsense reasoning, while its Yi-VL variant expands to multimodal abilities by integrating a Vision Transformer for image processing. Trained on a carefully curated 3.1 TB multilingual corpus, the model is designed for high-quality task performance, although it shares common LLM limitations such as potential hallucination and repetition issues.

Capabilities

Providers(5)