Yi (01.AI)

Definition

Yi is a family of large language models developed by 01.AI, a Chinese AI startup founded by Kai-Fu Lee, first released as open-weight models in 2023 and later extended with larger, more capable variants including Yi-Large.

Overview

01.AI positioned Yi as a strong open-weight alternative to Western open models like Llama, trained with a bilingual focus on both English and Chinese text, which made it particularly popular among developers building Chinese-language applications. The initial release included multiple sizes, most notably 6-billion and 34-billion parameter variants, made available with permissive-leaning licensing that encouraged broad adoption and fine-tuning by the open-source community. Following the initial open-weight releases, 01.AI expanded the family with larger, more capable models such as Yi-Large, offered through a hosted API rather than as open weights, aiming to compete more directly with closed frontier offerings on general capability rather than only on openness and efficiency. This mirrors a broader pattern among several Chinese AI labs of the era, releasing open-weight models to build developer mindshare while also offering larger proprietary models commercially. Yi models are commonly used as a base for further fine-tuning within the open-source Hugging Face ecosystem, and they're frequently cited as part of the broader wave of strong Chinese open-weight Foundation Models that emerged alongside models like GLM (Zhipu AI) and DeepSeek's model families.

Key Features

• Developed by 01.AI, a Chinese AI startup founded by Kai-Fu Lee
• Initial 2023 release included open-weight 6B and 34B parameter variants
• Trained with a strong bilingual focus on English and Chinese text
• Later extended with larger proprietary models such as Yi-Large via hosted API
• Popular base model for fine-tuning within the open-source community
• Part of the broader wave of competitive Chinese open-weight foundation models

Use Cases

Frequently Asked Questions

Related in AI Models