OpenChat 3.2

About

OpenChat 3.2 is an open-source large language model built on a transformer architecture, optimized for text input and excelling in both single-turn and multi-turn conversations. Developed through supervised fine-tuning (SFT) with 80,000 ShareGPT conversations, it achieves high performance despite its straightforward methods. Notably efficient, it is fine-tuned within 10 hours using 8xA100 80G hardware. Freely available for commercial use under the Llama 2 Community License, OpenChat 3.2 is known for generating human-quality text and engaging in natural, coherent dialogue across various topics. It ranks highly on benchmarks such as AlpacaEval and MT-bench among 13B parameter models.

Capabilities