Sure! Here’s the translation to American English:
As organizations expand their use of generative artificial intelligence, there is a growing need to process large volumes of data in a more cost-effective manner. In response to this demand, Amazon Bedrock has launched batch inference, a solution that enables mass data processing at a cost up to 50% lower than real-time inference. This tool is ideal for tasks such as analyzing historical data, large-scale text summarization, and other workloads that are not time-sensitive.
The launch of batch inference comes with significant enhancements in Amazon Bedrock, including increased support for models and optimizations that improve performance and cost transparency. Among the new available models are options like Claude Sonnet 4 from Anthropic and several models from OpenAI. These additions, along with performance improvements, enable greater efficiency in executing batch jobs.
Users can manage and monitor their batch inference jobs through Amazon CloudWatch, eliminating the need to develop custom monitoring solutions and ensuring complete visibility into the progress of tasks in their AWS accounts. Recommended use cases for this functionality include data analysis that can tolerate delays, knowledge base enrichment, and compliance reviews of sensitive content.
To initiate a batch inference job, users have multiple options: they can do so via the AWS Management Console, using AWS SDKs, or through the command line interface (CLI). This process is user-friendly and allows users to specify the necessary details to run the job, such as the model to be used and the locations of the input and output data.
Additionally, Amazon Bedrock now automatically provides metrics in the AWS/Bedrock/Batch namespace, enabling users to monitor the progress of their workloads. These metrics offer key insights into the status of jobs, including backlog size and overall performance.
With these enhancements, Amazon Bedrock not only optimizes batch inference but also equips organizations with tools to maximize the efficiency of their generative artificial intelligence workloads, encouraging them to effectively implement these innovations.
Source: MiMub in Spanish
