Introducing Llama 3.3: A Multilingual Language Model by Meta
meta-llama/Llama-3.3-70B-Instruct · Hugging Face 🔗
Llama 3.3 is a large language model developed by Meta, designed for various commercial and research applications, including multilingual dialogue. Users must accept the Llama 3.3 Community License Agreement, which details the terms for using, modifying, and redistributing the model. The model is optimized for safety and helpfulness, utilizing a unique training process that includes both human-generated and synthetic data. It supports multiple languages and is intended for responsible use, with strict guidelines against illegal and harmful activities. Meta emphasizes the importance of safety and ethical considerations in the deployment of Llama 3.3, urging developers to conduct proper testing and adhere to community standards.
- Release Date: December 6, 2024
- Model Size: 70 billion parameters
- Supported Languages: English, German, French, Italian, Portuguese, Hindi, Spanish, Thai
- Training Data: ~15 trillion tokens from publicly available sources
- Safety and Ethical Use: Strong emphasis on responsible deployment and adherence to community guidelines
What is the purpose of the Llama 3.3 model?
Llama 3.3 is designed for commercial and research use, focusing on multilingual dialogue and various natural language generation tasks.
What are the key guidelines for using Llama 3.3?
Users must comply with the Llama 3.3 Community License Agreement, which includes restrictions against illegal activities and mandates proper attribution when redistributing the model.
How does Meta ensure the safety of Llama 3.3?
Meta employs a multi-faceted approach to safety, including the use of fine-tuning data, ongoing evaluations, and community feedback to mitigate risks associated with the model's deployment.