MaziyarPanahi/calme-2.3-llama3-70b

TEXT GENERATIONPricing:Input $3.5 / Cached $0.7 / Output $8.3Concurrent Unit Cost:4Model Size:70BQuant:FP8Context Size:8kTool Calling:SupportedPublished:Apr 27, 2024License:llama3Architecture:Transformer0.0K Featherless Exclusive Cold

MaziyarPanahi/calme-2.3-llama3-70b is a 70 billion parameter language model fine-tuned by MaziyarPanahi using DPO from Meta-Llama-3-70B-Instruct, designed for general instruction-following tasks. It achieves an average score of 78.74 on the Open LLM Leaderboard, demonstrating strong performance across various benchmarks including MMLU (80.47) and GSM8k (87.19). This model is suitable for applications requiring a robust, instruction-tuned LLM with a context length of 8192 tokens.

Loading preview...

MaziyarPanahi/calme-2.3-llama3-70b Overview

This model is a 70 billion parameter language model, fine-tuned by MaziyarPanahi using Direct Preference Optimization (DPO) on the meta-llama/Meta-Llama-3-70B-Instruct base model. It is designed to excel in general instruction-following and conversational tasks, leveraging the strong capabilities of its Llama 3 foundation.

Key Capabilities

  • Instruction Following: Optimized through DPO for improved adherence to user instructions and preferences.
  • Strong Benchmark Performance: Achieves an average score of 78.74 on the Open LLM Leaderboard, indicating robust performance across diverse tasks.
  • Reasoning and Knowledge: Scores 80.47 on MMLU (5-Shot) and 72.35 on AI2 Reasoning Challenge (25-Shot), demonstrating strong multi-task language understanding and reasoning abilities.
  • Mathematical Proficiency: Achieves 87.19 on GSM8k (5-Shot), suggesting good performance in mathematical problem-solving.
  • Context Handling: Supports a context length of 8192 tokens, allowing for processing longer inputs and maintaining conversational coherence.

Good For

  • General-purpose chatbots and conversational AI: Its instruction-tuned nature makes it suitable for engaging in diverse dialogues.
  • Applications requiring strong reasoning and knowledge recall: Performance on MMLU and ARC indicates its utility for tasks needing broad understanding.
  • Educational tools and problem-solving assistants: Its GSM8k score suggests it can be effective in scenarios involving quantitative reasoning.
  • Developers seeking a high-performing Llama 3 variant: Offers an enhanced version of the Llama 3 70B Instruct model with DPO fine-tuning.

Finetunes
3 models