Zoho Corporation, a global technology company, has announced significant expansion of its artificial intelligence portfolio, unveiling its proprietary large language model (LLM) named Zia LLM, alongside a suite of AI-powered tools designed to enhance organisational workflows. The company claims that these innovations will deliver operational and financial efficiencies across various business functions by embedding AI into daily organisational tasks.
According to the firm, Zia LLM is entirely developed in-house using NVIDIA’s AI-accelerated computing platform and takes a multi-model approach with three configurations—1.3 billion, 2.6 billion, and 7 billion parameters—each tailored for specific contextual use cases such as structured data extraction, summarisation, retrieval-augmented generation, and code generation. Zoho asserts that this approach allows optimisation balancing computational resource management and functional power, benchmarking competitively against existing open-source models. The company also highlighted the imminent deployment of Zia LLM across its data centres in the US, India, and Europe, with availability to customers anticipated within months.
In addition to the LLM, Zoho introduced two proprietary Automatic Speech Recognition (ASR) models catering to English and Hindi, which reportedly outperform comparable models in accuracy while operating with low computational demands. Expansion to include other languages, particularly those spoken across Europe and India, is planned.
Building on these foundational AI models, Zoho is launching Zia Agents—an array of over 25 prebuilt AI agents accessible via the newly established Agent Marketplace. These AI-driven autonomous bots are embedded within Zoho’s suite of over 55 applications and designed to assist roles ranging from sales development and customer support to account management. Examples include the revamped Ask Zia conversational assistant tailored for various data roles, a Customer Service Agent for Zoho Desk that can handle and triage incoming queries, as well as specialised agents like Revenue Growth Specialist and Candidate Screener. Users are also empowered to craft their own agents through the Zia Agent Studio, a no-code, prompt-based agent builder with access to more than 700 actions across Zoho’s ecosystem. The platform also supports deployment versatility, from autonomous operation to integration within customer dialogues or rule-based activations.
Notably, Zoho has integrated a Model Context Protocol (MCP) server to facilitate interoperability, enabling third-party agents to access Zoho data and actions while enforcing established user permission structures, thereby promising seamless collaboration and extension beyond Zoho’s native applications.
The company emphasises a strong privacy stance, underscoring that all AI initiatives are developed internally to ensure data remains within Zoho’s secure servers and data centres, avoiding the need to send sensitive information to external AI cloud providers. This aligns with Zoho’s broader privacy commitment, operating without an ad-revenue model and owning its data centres to maintain control over privacy and security.
Zoho’s CEO, Mani Vembu, stated that the company’s goal is to democratise advanced AI technologies globally without sacrificing data privacy or organisational flexibility. Yet, while Zoho’s proprietary models and privacy-first approach may appeal to organisations concerned about data governance, the competitive AI landscape is rapidly evolving, with established players offering various third-party AI integrations. Zoho’s strategy to right-size models and maintain a balance between performance and resource efficiency seeks to differentiate its offerings from broader, sometimes more resource-intensive AI platforms.
The company plans ongoing enhancements, including scaling Zia LLM’s parameters by the end of the year, extending language support in speech recognition, and introducing reasoning language models. Future capabilities will also include enabling Zia Agents to interact with each other and agents on other platforms, signalling a push towards more collaborative, cross-agent AI ecosystems.
While Zoho has positioned its AI expansion as a leap forward for businesses seeking AI-driven efficiencies, independent assessments of model performance and user adoption during the early access phase will provide greater clarity on how these tools compare to other AI solutions in terms of effectiveness, cost, and integration ease.
General availability for the Zia AI suite—including the LLM, Zia Agents, Agent Studio, and MCP server—is anticipated by the end of 2025, with pricing details yet to be disclosed. The early access programme aims to gather usage data across industries and geographies to refine these offerings ahead of broad market release.
Source: Noah Wire Services
Subscribe to Industry Updates
