Zoho, the global technology company, has announced a major expansion of its AI offerings with the launch of Zia LLM, its proprietary large language model, alongside a no-code agent builder called Zia Agent Studio.
Built entirely in-house on NVIDIA’s AI-accelerated computing platform, Zia LLM is designed for enterprise scenarios such as structured data extraction, summarisation, retrieval-augmented generation (RAG), and code generation. The family includes three models—1.3B, 2.6B, and 7B parameters—each trained and optimised for contextual applicability and benchmarking competitively with comparable open-source options. This tiered approach lets Zoho match the right model to each use case, balancing performance and resource efficiency. Zoho says model sizes will scale over time, with the first parameter increases planned before the end of 2025.
“This announcement emphasises Zoho’s longstanding aim to build foundational technology focused on protection of customer data, breadth and depth of capabilities because of the business context, and value,” said Kehinde Ogundare, Country Head, Zoho Nigeria.
“Our LLM model is trained specifically for business use cases, keeping privacy and governance at its core, which has resulted in lowering the inference cost, passing on that value to the customers, while also ensuring that they are able to utilise AI productively and efficiently.”
While Zoho continues to support popular LLM integrations—such as ChatGPT, Llama, and DeepSeek—Zia LLM reinforces the company’s privacy posture by keeping customer data on Zoho servers, enabling organisations to leverage modern AI without sending data to external AI clouds. The model is currently being tested across Zoho’s app ecosystem and is slated for broader customer availability in the coming months.
Zoho also introduced two proprietary Automatic Speech Recognition (ASR) models for English and Hindi. Optimised for low compute loads without sacrificing accuracy, the models benchmark up to 75% better than comparable systems on standard tests. The company plans to expand language coverage to support more regions and will also introduce a reasoning language model (RLM).
To accelerate real-world adoption of agentic AI, Zoho has embedded a suite of AI agents directly within its products, enabling role-aware assistance across business workflows. Examples include a Customer Service Agent for Zoho Desk that processes incoming requests, understands context, and either resolves issues or routes them to human agents—serving as an efficient first line of support.
Ask Zia, Zoho’s conversational assistant available across the platform, now gains additional BI capabilities tailored to data engineers, analysts, and data scientists while remaining accessible to all users. It can build end-to-end data pipelines for engineers, analyse datasets, and generate reports and dashboards through interactive conversations for analysts, as well as jump-start ML model building for data scientists.
First announced earlier in 2025, Zia Agent Studio has been simplified to a fully prompt-based experience (with a low-code option) and provides ready access to more than 700 actions across Zoho’s products. User-built agents can run autonomously, be triggered by button clicks, follow rule-based automations, or be invoked within customer conversations, depending on operational needs.
On deployment, agents can be provisioned as “digital employees,” inheriting an organisation’s existing permission structures. Administrators can run behavioural audits and evaluate performance and impact, ensuring each agent operates within clearly defined guardrails and delivers measurable value.
Several pre-built agents are already available, including Candidate Screener for ranking applicants by role requirements and experience, Deal Analyser for estimating win probabilities and suggesting next best actions, and Revenue Growth Specialist for surfacing upsell and cross-sell opportunities. These can be discovered and deployed via the Agent Marketplace, with ecosystem partners, ISVs, and individual developers able to create and list their own agents in the coming months.
Zoho plans to extend Ask Zia’s skill set further, initially focusing on Finance and Customer Support functions. The company will also implement support for the Agent-to-Agent (A2A) protocol, enabling Zia Agents to collaborate with each other and integrate with agents on other platforms.
For interoperability and governance, Zoho has adopted the Model Context Protocol (MCP), allowing clients to securely access and interact with workflows and actions across more than 15 Zoho applications. MCP is in early access, and integrations via Zoho Flow extend its reach to third-party tools. Zoho Analytics also supports a local MCP server, enabling advanced contextual AI scenarios while maintaining strong data security and compliance.