Overview

Weyaxi/OpenHermes-2.5-neural-chat-7b-v3-1-7B is a 7 billion parameter language model developed by Weyaxi. It is a merge of two prominent models: teknium/OpenHermes-2.5-Mistral-7B and Intel/neural-chat-7b-v3-1, utilizing a ties merge approach. This combination aims to integrate the distinct capabilities of its constituent models.

Key Capabilities

• Instruction Following: Inherits strong instruction-following abilities from its base models.
• General Conversational AI: Designed for broad conversational applications.
• Performance: Achieves a competitive average score of 67.84 on the Open LLM Leaderboard, with specific scores including:
  • ARC (25-shot): 66.55
  • HellaSwag (10-shot): 84.47
  • MMLU (5-shot): 63.34
  • TruthfulQA (0-shot): 61.22
  • Winogrande (5-shot): 78.37
  • GSM8K (5-shot): 53.07

Good For

• Chatbot Development: Ideal for creating responsive and coherent conversational agents.
• Instruction-based Tasks: Suitable for applications requiring the model to follow specific commands or prompts.
• Research and Experimentation: Provides a robust base for further fine-tuning or architectural exploration, leveraging the combined strengths of its merged components.

Prompt Templates

The model supports multiple prompt templates, with ChatML (from OpenHermes-2.5-Mistral-7B) being recommended, alongside the template from neural-chat-7b-v3-1.

Quantized Versions

Optimized, quantized versions are available from TheBloke in GPTQ, GGUF, and AWQ formats for efficient deployment.