Zoho Corporation has recently unveiled 'Zia,' a proprietary large language model (LLM) designed to power its suite of applications and services. This announcement marks a significant step for the company in strengthening its artificial intelligence (AI) capabilities while maintaining a strong focus on data privacy. The introduction of Zia LLM is part of a broader initiative that includes new AI infrastructure investments, showcasing Zoho's commitment to providing cutting-edge technology without compromising organizational flexibility and data security.
Zia LLM has been developed entirely in-house using NVIDIA's AI accelerated computing platform. The model is specifically trained for Zoho product use cases, including structured data extraction, summarization, retrieval-augmented generation (RAG), and code generation. To cater to diverse business needs, Zia LLM comes in three parameter sizes: 1.3 billion, 2.6 billion, and 7 billion. Each model is optimized for different business contexts and benchmarked against similar open-source models to ensure competitive performance. This "right-sizing" approach allows Zoho to balance computational power with efficiency, optimizing performance based on user needs.
Alongside Zia LLM, Zoho is introducing several new AI tools and infrastructure components. These include over 25 prebuilt AI-powered agents available in an Agent Marketplace, the no-code Zia Agent Studio for custom agent building, and a Model Context Protocol (MCP) server that provides third-party agents access to Zoho's extensive library of actions. The Zia Agent Studio allows users to create their own AI assistants without coding, enabling them to automate tasks, respond to customer queries, and assist teams with analysis. Over 700 built-in actions are available across Zoho's product suite, making it easier for businesses to integrate AI into their workflows.
The MCP server is designed to enable interoperability with third-party clients, providing secure access to a library of actions from more than 15 Zoho applications. This allows for the creation of interoperable AI workflows, enhancing the functionality and reach of Zoho's AI ecosystem. Zoho also plans to expand Ask Zia, its conversational AI assistant, to serve finance and customer support teams, and enable collaboration between agents through an Agent2Agent protocol.
In addition to Zia LLM, Zoho is launching proprietary Automatic Speech Recognition (ASR) models capable of performing speech-to-text conversion in both English and Hindi. These models are optimized for accuracy and low computational load, showing up to a 75% improvement over comparable alternatives in standard benchmarks. Zoho plans to add support for more languages in the future, particularly those spoken in Europe and India.
Zoho emphasizes that Zia LLM allows customers to maintain their data on Zoho's servers, ensuring they retain control over privacy and security. This is a key differentiator, as it allows businesses to leverage the power of AI without transmitting sensitive data to external AI cloud providers.
The rollout of Zia LLM will occur across data centers in the United States, India, and Europe. Initially, it will support internal Zoho applications, with customer deployment expected in the coming months. Zoho has short-term and long-term plans for its AI ecosystem, including scaling Zia LLM's model sizes, expanding speech-to-text support for more languages, and releasing a reasoning language model (RLM).
