AWS "Offering Over 100 LLMs on Bedrock Market...Including Korean AI Models"

AI Agent Construction and Coordination
New Bedrock Features Also Announced

Amazon Web Services (AWS) has launched the AI platform 'Amazon Bedrock Marketplace' to expand the generative artificial intelligence (AI) ecosystem. AI models from Korean companies such as LG AI Research, NCSOFT, and Upstage are also included.

On the 4th (local time), Swami Sivasubramanian, Vice President of AWS AI and Data, announced at 'AWS re:Invent 2024' held in Las Vegas, USA, that "the Bedrock Marketplace provides access to more than 100 new models offered by major providers."

In the Amazon Bedrock Marketplace, customers can select and utilize AI models from around 100 large language models (LLMs) that suit their purposes. They can also use tools such as Bedrock’s guardrail agents (safety devices) and benefit from built-in security and privacy features.

The Bedrock Marketplace includes Mistral AI’s Mistral Nemo Instruct 2407, NVIDIA’s NIM microservices, and more. It is the first cloud provider to offer models from Luma AI and Poolside.

Among Korean companies, it supports LG AI Research’s ExaONE, NCSOFT’s Barco, and Upstage’s large language models (LLMs) SOLA Mini and SOLA Pro.

Vice President Sivasubramanian said, "Amazon Bedrock is rapidly growing by providing a wide range of model options, customized tools that make it easy to utilize customer data, responsible AI capabilities, and the ability to develop sophisticated agents."

New features of Amazon Bedrock were also announced. Amazon Bedrock is a fully managed service that enables easy building and scaling of generative AI applications using high-performance foundational models. These features include automated inference verification, a safety mechanism to prevent factual errors caused by AI hallucinations; multi-agent collaboration, which allows easy building and coordination of multiple AI agents; and model distillation, which transforms specific knowledge from large high-performance models into smaller, more efficient models.