In a major push to support deep-tech innovation, Google has announced significant expansions to its Gemma family of open AI models, unveiling new variants specifically tailored for healthcare applications and agent-based systems, two domains witnessing rapid adoption in India’s burgeoning AI ecosystem. The announcements came on Jan 15, 2026, at the Google AI Startups Conclave held this week, where the tech giant also introduced a broader Market Access Program for Indian AI startups.
Healthcare-Focused Gemma Models Aim to Scale Clinical AI
The introduction of MedGemma 1.5 and FunctionGemma signals Google’s broader effort to equip developers with open, purpose-built AI models that can be deployed at scale. MedGemma 1.5, a healthcare-focused model with four-billion parameters, is designed to support large-volume medical AI applications, ranging from medical imaging analysis to pathology workflows and clinical report interpretation. The model is an extension of Google’s partnership with the All India Institute of Medical Sciences (AIIMS) to develop India-specific Health Foundation Models, aimed at strengthening diagnostic capabilities and improving healthcare delivery across the country.
In parallel, Google introduced FunctionGemma, a lightweight model optimized for function calling that enables developers to create responsive, on-device AI agents. Unlike traditional models that depend on cloud connectivity, FunctionGemma can operate with minimal latency and without continuous internet access, making it especially useful for applications in regions with limited connectivity or for privacy-sensitive use cases. These agent-based systems can automate workflows, handle local decision-making, and integrate with external tools and services — a capability seen as essential for next-generation AI assistants and automation platforms.
Open Models as Building Blocks for Deep-Tech Startups
Google’s expansion of the Gemma model family reflects a broader trend in the AI sector: open models are becoming foundational to startup innovation, particularly in areas that require specialized domain expertise, such as health and mobile AI. The company’s open-weight Gemma models, developed by Google DeepMind, have been iteratively expanded since their first release in early 2024 and include a suite of variants tailored for different performance, modality, and deployment needs.
Market Access Program Complements Model Expansion
The announcements come at a time when India’s AI market is projected to reach over $120 billion by 2030, according to industry reports, with nearly half of enterprises already moving from pilot projects to production-ready applications. Google’s Market Access Program announced at the same event aims to help startups bridge the gap between prototype and scale. It provides curriculum support, guidance on enterprise sales strategies, and access to Google’s global network of CIOs and CXOs, helping emerging AI companies go global.
Positioning Gemma for India’s Next Wave of AI Innovation
Preeti Lobana, Vice President and Country Manager for India at Google, underscored the company’s commitment to deep-tech ecosystems: “AI startups are no longer experimenting at the edges — they’re turning capability into products that people use, trust and pay for.” The new Gemma models, she added, will equip developers with practical, scalable tools to address real-world challenges in healthcare, automation, and beyond. By combining open-model innovation with robust startup support programs, Google is positioning itself at the center of India’s next wave of AI development, one where domain-specific intelligence and agentic AI systems play a critical role in driving economic and social impact.
