Kai-Fu Lee’s AI Company “01.AI” Announces the Open Source of the Yi-9B Model

On March 6th, “01.AI” official WeChat account announced the open sourcing of Yi-9B model. The official referred to it as the “top scorer in science” among the Yi series models – Yi-9B is currently the most powerful model in terms of code and mathematical capabilities within the Yi series models, with actual parameters at 8.8 billion and a default context length of 4K tokens.

The model is based on Yi-6B (trained with 3.1T tokens) and further trained with 0.8T tokens, data up to June 2023.

According to the introduction, in terms of overall performance (Mean-All), Yi-9B’s performance is “the best among open-source models of similar size”, surpassing DeepSeek-Coder, DeepSeek-Math, Mistral-7B, SOLAR-10.7B and Gemma-7B.

In terms of code capability (Mean-Code), Yi-9B’s performance is second only to DeepSeek-Coder-7B, outperforming Yi-34B, SOLAR-10.7B, Mistral-7B and Gemma-7B.

In terms of mathematical ability (Mean-Math), the Yi-9B is second only to the DeepSeek-Math-7B, outperforming the SOLAR-10.7B, Mistral-7B and Gemma-7B.

In terms of common sense and reasoning ability (Mean-Text), the Yi-9B’s performance is on par with the Mistral-7B, SOLAR-10.7B, and Gemma-7B.

The official statement also mentioned that Yi-9B (BF 16) and its quantized version Yi-9B (Int8) can be easily deployed on consumer-grade graphics cards, with low usage costs and developer-friendly features.

“01.AI” was founded by Kai-Fu Lee, Chairman and CEO of Sinovation Ventures. It has previously released two open-source large models, Yi-34B and Yi-6B, which are claimed to be completely open for academic research and simultaneously available for free commercial applications.

SEE ALSO: AI Expert Kaifu Lee Celebrated for His Contributions to Tech Development