Amazon Bedrock now supports import of custom models to access as a fully managed API

Amazon Web Services (AWS) today announced new Amazon Bedrock features and models that will allow developers to create better AI applications and experiences.

With the new Amazon Bedrock Custom Model Import support, organizations can now import their own custom LLM models and access them via a fully managed API in Bedrock. Organizations can add custom models to Amazon Bedrock from Amazon SageMaker, or via another third-party tool or cloud provider. This new Amazon Bedrock Custom Model Import is available now in preview and supports the following model architectures

• Flan-T5
• Llama
• Mistral

Amazon today announced the general availability of Amazon Titan Image Generator. Also, the images produced using Titan Image Generator now includes invisible watermarking. Amazon also announced that the latest version of Amazon Titan Text Embeddings, Amazon Titan Text Embeddings V2, which is optimized for Retrieval Augmented Generation (RAG) use cases. The Amazon Titan Text Embeddings V2 will allow customers to have flexible embedding sizes to reduce overall storage up to four times, while retaining 97% of the accuracy for RAG use cases. This new text embeddings model will be available later this month.

Along with the above announcements, AWS also announced the general availability of Guardrails for Bedrock and Model Evaluation.