Meta-Llama-3.1-405B-Instruct-FP8
Multi language dialogue generation model
- Supports text generation in 8 languages: English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai.
- Adopting an optimized Transformer architecture, adjustments are made through supervised fine-tuning and human feedback reinforcement learning.
- Provide a static model trained on data from December 2023, and continuously improve model security based on community feedback in the future.
- Comply with the custom commercial license Llama 3.1 Community License, detailed license agreement can be viewed on GitHub.
- Supports commercial and research purposes, suitable for various natural language generation tasks such as assistant chat and code generation.
- The training utilized 39.3M GPU hours of computing resources, and Meta has achieved net zero greenhouse gas emissions from global operations since 2020.
Product Details
The Meta Llama 3.1 series models are a set of pre trained and instruction adjusted multilingual large language models (LLMs), including models of three scales: 8B, 70B, and 405B, optimized for multilingual conversation use cases, with performance superior to many open source and closed source chat models.
