Zoho Corporation, a leader in technology, proudly introduces its proprietary large language model Zia LLM, along with 25+ AI-powered Zia Agents and the no-code Zia Agent Studio. These tools, combined with the innovative Model Context Protocol (MCP) server, present vast opportunities for third-party integration into Zoho's library.
These developments are designed to unlock the full potential of contextual and agentic AI technology in organizations, positively impacting daily workflows across various sectors. With these advancements, Zoho ensures substantial operational and financial improvements for businesses globally.
"Today's announcement underscores our commitment to building foundational technology that safeguards customer data, offers comprehensive capabilities, and ensures value," commented Sachin Agrawal, Managing Director of Zoho UK. Developed internally, Zoho's AI initiatives promise cutting-edge tools without compromising data privacy.
Constructed entirely in-house using NVIDIA's AI platform, Zia LLM is a testament to Zoho's innovation. This model encompasses three versions, each optimized for different use cases: structured data extraction, summarization, RAG, and code synthesis. The models, boasting parameters of 1.3 billion, 2.6 billion, and 7 billion, are tuned for precise contextual relevance. This strategic scaling ensures tailored solutions for varied user needs while balancing computational efficiency.
Alongside Zia LLM, Zoho introduces its Automatic Speech Recognition (ASR) models capable of English and Hindi speech-to-text conversion. These models outperform competitors by up to 75%, ensuring accuracy with minimal computational demand.
While Zoho remains inclusive of multiple LLM integrations—encompassing giants like ChatGPT and Llama—Zia LLM reinforces its dedication to user data privacy by securely hosting data on Zoho's servers.
Anticipated to roll out across data centers in the US, India, and Europe, Zia LLM's initial implementation aims at Zoho's expansive app suite, with preparations underway for broader customer access.
In the near future, Zoho plans to enhance Zia LLM with parameter expansions, expected by 2025's end. Furthermore, the addition of new language capabilities, catering to European and Indian dialects, and the launch of a reasoning language model (RLM) are on the horizon.
Complementary functions, such as Ask Zia, will evolve with additional skills to assist finance and customer support teams. Moreover, the forthcoming implementation of the Agent2Agent (A2A) protocol aims to facilitate seamless collaboration across diverse AI agents.
