Yi AI

The YI series models represent the next generation of open-source large language models, trained from the ground up. Designed as bilingual language models, the YI series has undergone training with a multilingual corpus exceeding 3T tokens, establishing itself as one of the most powerful LLMs globally. It demonstrates promising capabilities in language comprehension, commonsense reasoning, reading comprehension, and more. The Yi-34b-chat model ranks second (following GPT-4 Turbo) on the AlpacaEval leaderboard, surpassing other LLMs such as GPT-4, Mixtral, and Claude (based on data as of January 2024). The YI-34B model leads all existing open-source models (e.g., Falcon-180B, Llama-70B, Claude), achieving top rankings in both English and Chinese across various benchmarks, including the Hugging Face Open LLM Leaderboard (pre-training) and C-Eval (based on data available as of November 11, 2023). Credit goes to the open-source communities of Transformer and LLaMA, as they have reduced the effort required for building from scratch and enabled the use of shared tools within the AI ecosystem.

Yi AI is a series of open-source large language models trained from scratch by 01.AI, offering robust bilingual understanding, commonsense reasoning, and reading comprehension, excelling in global benchmarks.