unsloth/Llama-3.3-70B-Instruct

Model Overview

unsloth/Llama-3.3-70B-Instruct is a 70 billion parameter multilingual large language model (LLM) from Meta, designed for instruction-tuned generative tasks. It leverages an optimized transformer architecture with Grouped-Query Attention (GQA) for enhanced inference scalability and features a substantial 128k context length. The model is instruction-tuned using supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety.

Key Capabilities

• Multilingual Dialogue: Optimized for conversational use cases across English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai.
• Robust Performance: Achieves strong results on industry benchmarks, including 68.9% on MMLU Pro, 88.4% on HumanEval (pass@1), and 77.0% on MATH (sympy_intersection_score).
• Extensive Training: Pretrained on over 15 trillion tokens of publicly available online data, with a knowledge cutoff of December 2023.
• Tool Use Support: Integrates with various tool use formats, enabling advanced function calling capabilities.

Intended Use Cases

This model is suitable for commercial and research applications requiring assistant-like chat functionalities in multiple languages. Its capabilities also extend to leveraging model outputs for synthetic data generation and distillation, making it versatile for various natural language generation tasks. Developers can fine-tune the model for additional languages, provided they adhere to the Llama 3.3 Community License and Acceptable Use Policy.