Revolutionary AI Unveils Cutting-edge Yi Model, Approaches GPT-3.5 with 200,000-Token Data Support.

01. AI, a Chinese artificial intelligence company founded by Kai-Fu Lee, former Google executive, has released the latest LLM model named Yi with significant improvements, especially in handling large-scale data. Previously limited to a maximum of 32,000 tokens, Yi now supports up to 200,000 tokens as announced earlier.

Since its initial launch, Yi has had several minor versions released, including chatbot models, a 9B model, and an image-supporting model.

Previously criticized for utilizing the LLaMA 2 framework without sufficient credit, 01.AI now acknowledges its use of the LLaMA 2 structure. However, they emphasize that both models utilize the same Transformer technology, and they have retrained all models from scratch without relying on previously trained LLaMA 2 models. Additionally, the team has created a new high-quality dataset to develop efficient smaller models.

Source: GitHub: 01-ai/Yi

TLDR: 01.AI, led by Kai-Fu Lee, introduces the advanced LLM model Yi with enhanced data handling capabilities, overcoming previous criticisms by retraining models independently and creating a new dataset for efficient performance.