Consolidating its position as one of the leading platforms for custom-generated AI models, AWS announced the launch of custom model import within Bedrock, its suite of GenAI services for enterprises.
As the name suggests, custom model import allows organizations to import and access their own generative AI models as fully managed APIs. This means you can benefit from the same infrastructure and tools available with Bedrock's existing models.
AWS believes that the addition of support for custom models addresses the growing trend among companies to develop and refine their own in-house models.
AWS Bedrock now supports custom in-house models
Companies that use Bedrock's custom model import will not only be able to leverage their own models, but will also be able to use other tools in the suite for knowledge expansion, fine-tuning, protection against bias, and more. In theory, you should be able to offer your customers the best of both worlds.
Users can also monitor and filter output for unwanted content such as hate speech and violence, as well as evaluate model performance across a variety of criteria.
The service is currently available in preview and supports three of the most popular open model architectures: Flan-T5, Llama, and Mistal. AWS is working on adding more support for the architecture in the future.
At the same time, AWS also announced the general availability of Titan Image Generator and the launch of Titan Text Embeddings V2. The company says that training with AWS will reduce storage and compute costs while increasing accuracy, making the Titan family even more cost-effective for companies that don't have the resources to develop their own models. I am.
The Meta Llama 3 base model has also arrived at Bedrock, and Cohere's Command R and Command R+ models are expected to arrive soon as well.
Overall, it's refreshing to see AWS committed to interoperability by providing support for popular third-party and now in-house models on its platform. This is a move that could potentially set up AWS and its customers to jointly benefit.
