OpenOrca

About

The OpenOrca family of large language models (LLMs) leverages the Mistral 7B base model, fine-tuned with the OpenOrca dataset. This comprehensive dataset, mirroring the Orca Research Paper dataset, includes approximately 1 million GPT-4 completions and 3.2 million GPT-3.5 completions. OpenOrca models excel in a range of natural language processing tasks such as text and code generation, question answering, and conversations. Among these, the Mistral-7B-OpenOrca stands out for its superior performance among models with less than 30 billion parameters, achieving 98% of Llama2-70B-chat's performance on the HuggingFace leaderboard at its release. Designed to run efficiently on consumer-grade GPUs, these models are valuable tools for developers, though users should be mindful of potential biases inherent to the LLMs' training data1246810.