Search AI Products and News

Lingyi Wanwu officially released and open-sourced the fine-tuned model Yi-34B-Chat. Lingyi Wanwu provides 4-bit/8-bit quantized versions of the Yi-34B-Chat model, which has performed excellently in various evaluations. The Yi series models have been open-sourced in the ModelScope community, where developers can download, infer, and train these models.

["Orion Star has released the open source OrionStar-Yi-34B-Chat model based on the Zero One Thousand Things initiative, available for free commercial use.","The model performs excellently on the C-Eval and C-MMLU Chinese evaluation benchmarks.","OrionStar-Yi-34B-Chat is one of the outstanding open source alternatives to ChatGPT.","The model boasts high performance, is fully open-source, and is free for commercial use.","Orion Star is a provider of AI service-oriented solutions."]

The large model Yi-34B uses the LLaMA architecture but has changed tensor names, leading to controversy. Model performance includes a context window size exceeding 200k, processing 400,000 Chinese characters at once, and achievements in the Hugging Face Global Open Source Model Rankings. This has led the community to question whether Yi-34B has substantively altered the code and whether there is false advertising and license violations. The Zero One Everything company responded that the changes were based on the GPT structure, and that training requirements led to the renaming of code. The controversy focuses on whether the model violates open source agreements and the transparency of performance parameters and code modifications.

Kai-Fu Lee founded the startup Zero One Infinity, focusing on the development of large language models in China. The company released its first open-source model, Yi-34B, which has 34 billion training parameters. Zero One Infinity is valued at $1 billion and has received support from multiple investment institutions, experiencing rapid growth.