Overview

Groq/Llama-3-Groq-70B-Tool-Use is a 70 billion parameter causal language model, specifically fine-tuned by Groq for advanced tool use and function calling. It leverages the Llama 3 base model and has undergone full fine-tuning and Direct Preference Optimization (DPO) to enhance its capabilities in interacting with external tools and APIs.

Key Capabilities

• Advanced Tool Use: Designed to excel at tasks requiring API interactions, structured data manipulation, and complex function calling.
• High Performance: Achieves an overall accuracy of 90.76% on the Berkeley Function Calling Leaderboard (BFCL), positioning it as a top-performing open-source 70B LLM for tool use.
• Optimized Architecture: Built on an optimized transformer architecture, focusing on efficient and accurate tool integration.

Usage Considerations

• Specialized Focus: While highly effective for tool use, for general knowledge or open-ended conversational tasks, a general-purpose LLM might be more suitable.
• Sampling Sensitivity: The model is sensitive to temperature and top_p settings; recommended starting points are temperature=0.5 and top_p=0.65.
• Ethical Use: Inherits ethical considerations from the base Llama 3 model, requiring responsible implementation and additional safeguards as needed.