d-matrix/Llama-3.1-8B-Instruct
The d-matrix/Llama-3.1-8B-Instruct is an 8 billion parameter instruction-tuned large language model developed by Meta, part of the Llama 3.1 collection. Optimized for multilingual dialogue use cases, it features a 128k token context length and excels in general reasoning, code generation, and mathematical tasks. This model is designed for commercial and research applications, supporting multiple languages including English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai.
Loading preview...
Overview of Llama 3.1 8B Instruct
This model is the 8 billion parameter instruction-tuned variant from Meta's Llama 3.1 family, released on July 23, 2024. It leverages an optimized transformer architecture, fine-tuned using supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) for enhanced helpfulness and safety. Pretrained on over 15 trillion tokens with a knowledge cutoff of December 2023, it features a substantial 128k token context length and Grouped-Query Attention (GQA) for improved inference scalability.
Key Capabilities
- Multilingual Dialogue: Optimized for assistant-like chat in English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai.
- Enhanced Performance: Shows improvements over Llama 3 8B Instruct across various benchmarks, including MMLU (69.4%), HumanEval (72.6% pass@1), and MATH (51.9% final_em).
- Tool Use: Significantly improved tool-use capabilities, scoring 82.6% on API-Bank and 76.1% on BFCL, with support for advanced function calling.
- Code Generation: Strong performance in coding tasks, achieving 72.6% on HumanEval pass@1 and 72.8% on MBPP++.
- Reasoning & Math: Demonstrates robust reasoning (83.4% on ARC-C) and mathematical problem-solving (84.5% on GSM-8K).
Good for
- Commercial and Research Applications: Suitable for a wide range of natural language generation tasks and assistant-like chat.
- Multilingual AI Systems: Ideal for applications requiring robust performance across its 8 explicitly supported languages.
- Agentic Systems: Designed to be integrated into AI systems with additional safety guardrails, supporting tool use and external integrations.
- Synthetic Data Generation: Can be used to generate synthetic data and for model distillation to improve other models.