Fractal Analytics, a leading AI research company, has proposed to develop India’s first large reasoning model (LRM), a critical step toward advanced AI capabilities. The project, valued at ₹118.8 crore, is seeking ₹76.6 crore in funding from the IndiaAI Mission, an initiative by the Ministry of Electronics and Information Technology (MeitY).
This initiative marks a shift in India’s AI strategy. Unlike traditional pre-trained AI models, reasoning models focus on problem-solving, logical reasoning, and decision-making, making AI more context-aware and capable of handling complex real-world scenarios.
A Leap Beyond Pre-Trained AI Models
The global AI landscape is rapidly evolving, with post-training AI models gaining prominence. According to Fractal’s co-founder, Srikanth Velamakanni, the era of pre-training is over. AI research is now shifting toward models that can think, reason, and plan rather than just generate responses based on patterns.
Fractal’s proposed LRM series will include three distinct models:
A small model (2-7 billion parameters)
A medium model (20-32 billion parameters)
A large state-of-the-art model (70 billion parameters)
The project aims to process up to one trillion training tokens, positioning India’s AI ecosystem alongside leading nations in AI research.
India’s Push Toward AGI and AI Sovereignty
The global AI race is intensifying, with China’s DeepSeek R1 model currently leading with 671 billion parameters. Fractal’s initiative is seen as a crucial step for India to bridge the AI gap with global leaders like the US and China and advance toward Artificial General Intelligence (AGI).
The IndiaAI Mission has already received 187 applications for AI foundation models, with MeitY planning to approve select projects by the end of the month. Fractal’s proposal is among the most ambitious, as it seeks to develop reasoning capabilities from scratch using open-source large language models (LLMs) with permissive licenses.
Also read: Microsoft at 50: AI and Cloud Drive Future Growth
Building a World-Class Dataset for AI Reasoning
Fractal’s project aims to enhance AI’s logical reasoning capabilities by training models on real-world datasets from India’s top competitive exams. The company plans to create a diverse AI dataset sourced from leading examinations, including:
JEE Advanced (engineering)
NEET-PG (medical)
National Olympiads (STEM disciplines)
CAT and GATE (business and technology)
By leveraging these datasets, Fractal seeks to build an AI system that excels in STEM education, coding, medical decision-making, and agentic AI systems.
The Future of AI: Scaling Beyond Training to Reasoning
AI research has traditionally focused on scaling train-time compute, but performance plateaus and data saturation have led to diminishing returns. Fractal argues that scaling inference-time compute—which focuses on AI’s ability to reason, rather than just memorize—will offer a cost-optimal path to AGI.
This approach positions India as a leader in next-generation AI systems, focusing on logical reasoning, structured decision-making, and advanced planning capabilities. If approved, this project could transform India’s AI research landscape, giving Indian startups and enterprises access to world-class AI models designed for reasoning-based applications.
What’s Next for India’s AI Mission?
The IndiaAI Mission has emerged as a key government initiative to develop sovereign AI foundation models and reduce dependence on foreign AI technologies. With Fractal’s LRM proposal, India has the potential to accelerate AI-driven problem-solving and business intelligence.
As the AI landscape shifts toward post-training models, Fractal’s project could set a new benchmark in India’s artificial intelligence ecosystem, bringing AI-driven innovation in business, education, and research. The coming weeks will determine whether this ambitious proposal gets the green light from MeitY, potentially marking a new era for India’s AI leadership on the global stage.
