Brand Makers
Dil Ka Jod Hai, Tootega Nahin
PhysicsWallah launches Aryabhata AI model for precision-based JEE preparation
Aryabhata was trained on 130,000 high-quality question-answer pairs sourced from PW’s internal content bank.
Aryabhata was trained on 130,000 high-quality question-answer pairs sourced from PW’s internal content bank.
Edtech major PhysicsWallah (PW) has introduced Aryabhata 1.0, a purpose-built small language model (SLM) designed to support students preparing for the mathematics section of the JEE Main exam. The open-source model, named after the legendary Indian mathematician Aryabhata, is engineered with 7 billion parameters and delivers precise, exam-style responses tailored for competitive test environments.
Unlike general-purpose AI models, Aryabhata 1.0 is hyper-focused on academic rigor and problem-solving accuracy, reflecting India’s deep-rooted legacy in mathematical innovation. PW is making the model freely accessible to educators, developers, and researchers, encouraging widespread experimentation and innovation in education technology.
Outperforming the Giants Early benchmark results are promising. Aryabhata 1.0 scored 90.2% on the JEE Main April 2025 papers, outperforming major AI models like OpenAI’s GPT-4o (43.55%) and GPT-4.1 (80.44%). On reasoning and math evaluation datasets like Math 500 and GSM8K, the model achieved scores of 83.6% and 94.84%, respectively.
Despite being smaller than most large language models (LLMs), Aryabhata demonstrates high efficiency, functioning optimally within a 2,000-token window, compared to the 8,000-token limits commonly required by larger models.
Engineered for Education Aryabhata was trained on 130,000 high-quality question-answer pairs sourced from PW’s internal content bank. It uses a training methodology that blends supervised fine-tuning, reinforcement learning, and rejection sampling to align its outputs with the specific format and logic of competitive exams. This educational grounding makes it more intuitive and relevant for students tackling real-world exam scenarios.
Aryabhata 2.0 and Beyond The launch of Aryabhata 1.0 is the first milestone in PW’s broader AI strategy. The upcoming Aryabhata 2.0—expected later this year—will expand subject coverage to Physics and Chemistry and extend support to other key exams like JEE Advanced, NEET, and foundational learning levels.
This rollout is part of PW’s ongoing partnership with Microsoft Research India, which forms a key pillar of Microsoft’s $3 billion initiative aimed at strengthening India’s AI infrastructure. Together, the organisations are working on developing education-first AI models that boost tutoring accuracy and learning efficiency.
AI's Growing Footprint in Education Aryabhata’s launch underscores the rising relevance of AI in India’s test-prep landscape. A recent report by ByteDance noted that advanced models like Google’s Gemini 2.5 Pro are already outperforming top human scorers on IIT-JEE Advanced—signaling that AI could soon reshape the competitive exam experience entirely.
SPOTLIGHT
"The raucous, almost deafening, cuss words from the heartland that Piyush Pandey used with gay abandon turned things upside down in the old world order."
Read MoreThe new face of the browser: Who’s building AI-first browsers, what they do and how they could upend advertising
From OpenAI’s ChatGPT-powered Atlas to Microsoft’s Copilot-enabled Edge, a new generation of AI-first browsers is transforming how people search, surf and interact online — and reshaping the future of digital advertising.