IPO-bound PhysicsWallah unveils Aryabhata 1.0 AI model to help JEE aspirants

This AI initiative also comes at a time when PW is preparing for a public market listing

IPO-bound edtech unicorn PhysicsWallah (PW) has announced the launch of Aryabhata 1.0, a small language model (SLM) built specifically for mathematics preparation for engineering entrance exams like the JEE Main.

The model, revealed by co-founder Prateek Maheshwari in a social media post, marks the company’s early step into building exam-focused AI tools for Indian learners.

Despite being compact, Aryabhata 1.0 is built on a 7-billion parameter model and has shown strong performance in solving past JEE Main papers. According to Maheshwari, it scored 86 percent in the January 2024 session and 90.2 percent in April, outperforming much larger models while being trained on just a single H100 GPU.

The system was powered by 1,30,000 curated question-and-answer pairs and fine-tuned using techniques such as supervised fine-tuning (SFT), reinforcement learning (RL), and rejection sampling.

“Sometimes, the right support at the right moment can change a student’s entire path,” Maheshwari wrote. “At PW, we believe AI can offer that support if it’s built with care, context, and purpose.” He added that the startup is now working to expand Aryabhata 1.0’s capabilities to JEE Advanced and other mathematical domains.

Aryabhata 1.0 represents PW’s broader ambition to build subject-specific AI tools grounded in pedagogy rather than generic large language models. It also comes amid rising interest in AI applications for education, with competitors also exploring tutoring chatbots and adaptive learning.

The company has not yet disclosed whether Aryabhata 1.0 will be integrated into existing products or offered as a standalone service.

This AI initiative also comes at a time when PW is preparing for a public market listing. In March, the firm filed draft IPO papers with capital markets regulator Sebi via a confidential pre-filing route.

The confidential pre-filing route allows the company to withhold public disclosure of details under the draft red herring prospectus (DRHP). If the listing plans fructify, the firm could potentially become India's first edtech player to debut on the domestic bourses.

Noida-headquartered PhysicsWallah counts teacher-turned-entrepreneur Alakh Pandey and Prateek Maheshwari as co-founders. The firm has over 55 lakh paid students, 4.6 crore YouTube channel subscribers, 14,000-plus employees and covers at least 105 cities with 2 lakh plus students in offline centres.

In September, PhysicsWallah announced raising $210 million via a combination of primary and secondary transactions. The capital raise, led by Hornbill Capital, saw participation from Lightspeed Venture Partners, as well as existing investors WestBridge Capital and GSV Ventures.