Salesforce has announced the expansion of its large action models, enhancing their accessibility and flexibility in deployment across various environments. The software vendor has introduced model variations designed specifically for on-device implementation, use in environments with limited GPU resources, and industrial applications. The new set of models, called xLAM, is engineered to predict and perform the next actions to support AI agents.
These models are optimized to be smaller than typical large language models (LLMs), which results in lower costs, faster inference times, and improved sustainability. One of the standout features of the new models is their support for multi-turn tool calling. This capability allows AI agents to ask clarifying questions and adapt actions based on new information, which boosts their functionality and reliability in enterprise settings.
Salesforce aims to address several challenges in the deployment of AI agents.
Enhanced AI models boost reliability
According to Silvio Savarese, chief scientist at Salesforce AI Research, current LLM intelligence is often inconsistent and unreliable for enterprise deployment.
Shelby Heinecke, senior AI research manager at Salesforce, added that while significant progress has been made, much work remains to develop highly capable and consistent AI agents. To tackle these issues, Salesforce has developed new benchmarks and frameworks to test and evaluate models and agents more effectively. The company has also introduced measures to defend against prompt injection attacks and enhance toxicity detection, addressing security concerns critical for enterprises considering AI agents.
Security in AI applications remains a significant concern for enterprises. A report from Cloudera emphasized the need for stronger data privacy and security features in agent tools. Gartner predicts that approximately one in four enterprise breaches over the next three years could be attributed to AI agent abuse, underlining the urgent need for robust security measures.
In the evolving landscape of AI, Salesforce’s advancements in large-scale models represent a significant step toward making AI agents more reliable and accessible for enterprises.
Image Credits: Photo by Alex Kotliarskyi on Unsplash
Noah Nguyen is a multi-talented developer who brings a unique perspective to his craft. Initially a creative writing professor, he turned to Dev work for the ability to work remotely. He now lives in Seattle, spending time hiking and drinking craft beer with his fiancee.
