Anthropic has launched fine-tuning capabilities for Claude 3 Haiku on Amazon Bedrock. The feature is available for preview in the US West (Oregon) AWS region, allowing businesses to customise the Claude 3 model for specific tasks.
Claude 3 Haiku, which is a part of the Claude 3 family alongside Sonnet and Opus, can now undertake tasks like classification, API interactions, and data interpretation. The fine-tuning process uses prompt-completion pairs to enhance the model’s performance in specialised areas. The entire process is guided through the Amazon Bedrock console or API, where users can test and refine their custom models before deployment.
Several enterprise customers have already begun using the fine-tuning feature, allowing customisation of Haiku to fit their needs. SK Telecom reported a 73% increase in positive feedback for agent responses and a 37% improvement in key performance indicators after implementing a fine-tuned Claude model. Thomson Reuters plans to fine-tune Claude 3 Haiku with their industry expertise.
The features serves as a boon to enterprise customers, as it gives them more control over the model to train it according to their needs. Specifically, allowing the fine-tuning of Haiku will help in improving accuracy in specialised tasks, allow for faster processing at lower costs, give enterprises consistent output formatting, make the API accessible for companies, and ensuring data security for AWS customers.
As AI technology advances, customisation options like fine-tuning have become more abundant, especially as offerings from cloud infrastructure providers. This development in AI customisation could impact various sectors, from customer service to data analysis, as businesses seek to integrate AI more deeply into their operations.
Additionally, other cloud infrastructure companies like Microsoft Azure and GCP have also increased their GenAI offerings to businesses, giving them the option to build agents on the models they offer.
