Amazon Web Services (AWS) announces the availability of the Llama3.1 model on Amazon Bedrock, a series of advanced and powerful AI models developed by Meta. The Llama3.1 models include versions with 8B, 70B, and 405B parameters, demonstrating state-of-the-art performance on a wide range of industry benchmarks and offering new capabilities for generative AI applications.
All Llama3.1 models support a context length of 128K, which is 16 times the capacity of the Llama3 models, and are tailored for multilingual dialogue use cases in eight languages, including English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai, enhancing inference capabilities. The specific models include:
Llama3.1405B: This is currently the largest publicly available large language model, suitable for enterprise-level applications and research, particularly adept at general knowledge, long-form text generation, multilingual translation, machine translation, coding, mathematics, tool use, enhanced context understanding, and advanced reasoning and decision-making.
Llama3.170B: Ideal for content creation, conversational AI, language understanding, research, and enterprise applications, it excels in text summarization and accuracy, text classification, sentiment analysis and nuanced reasoning, language modeling, dialogue systems, code generation, and following instructions.
Llama3.18B: Designed for scenarios with limited computational power and resources, it specializes in text summarization, text classification, sentiment analysis, and language translation with low-latency inference requirements.
Meta has tested the performance of Llama3.1 on over 150 benchmark datasets covering a wide range of languages and conducted extensive human evaluations. Llama3.1 outperforms Llama3 in each major benchmark category.
Amazon Bedrock integrates responsible AI features with Llama3.1, providing data governance and model evaluation capabilities to help users confidently build secure and reliable generative AI applications.
Amazon Bedrock Guardrails allows users to create custom safeguards, promoting secure interactions, and continuously monitoring and analyzing user inputs and model responses.
The model evaluation feature on Amazon Bedrock enables users to select the best Llama model through automated or human evaluations.