OrionStarAI/OrionStar-Yi-34B-Chat-Llama
OrionStarAI/OrionStar-Yi-34B-Chat-Llama is a 34 billion parameter chat model developed by OrionStarAI, based on the Yi-34B architecture. Fine-tuned on over 15 million high-quality sentences, it excels in interactive chat experiences and demonstrates strong performance across Chinese, English, and general domain benchmarks. This model is optimized for general conversational AI tasks, offering an open-source alternative for ChatGPT-like applications.
Loading preview...
Model Overview
OrionStar-Yi-34B-Chat-Llama is a 34 billion parameter chat model from OrionStarAI, built upon the Yi-34B foundation. It has been extensively fine-tuned using a high-quality corpus exceeding 15 million sentences to enhance its interactive capabilities and general performance. This model is designed to provide an excellent conversational experience, particularly in the context of open-source alternatives to models like ChatGPT.
Key Capabilities & Performance
The model demonstrates impressive performance across various benchmarks, including C-Eval, MMLU, and CMMLU, covering Chinese, English, and general domains. Notably, OrionStar-Yi-34B-Chat achieves 77.71 on C-Eval, 78.32 on MMLU, and 73.52 on CMMLU, outperforming several larger models like LLaMA-2-70B-Chat and Baichuan2-13B-Chat in these specific evaluations. The LLaMA-formatted tensors facilitate automatic evaluation on the Hugging Face leaderboard.
Intended Use Cases
- General Conversational AI: Ideal for developing interactive chat applications and virtual assistants.
- Multilingual Applications: Strong performance in both Chinese and English contexts makes it suitable for diverse linguistic tasks.
- Academic Research: Fully open for academic research, adhering to the specified licenses.