OpenHermes
About
The OpenHermes family of large language models (LLMs) are built upon the Mistral foundation model, renowned for their stellar performance on benchmarks involving reasoning, instruction adherence, and providing truthful outputs 458. Fine-tuned on a substantial dataset with contributions from GPT-4 generated content and other high-quality open-source data, these models leverage ChatML for superior multi-turn conversation handling 4511. They offer various quantized versions, balancing resource needs and performance 238. OpenHermes models are recognized for their human-like text generation and strong instruction following, although they can be verbose 11. The OpenHermes-2.5-Mistral-7B model stands out for significant improvements and competitive edge over other Mistral adaptations 45.