HT Correspondent
TEZPUR, Nov 15: MK3, a multilingual artificial intelligence system developed entirely in India has demonstrated remarkable predictive accuracy in its early forecast of the 2025 Bihar Assembly elections. Powered by the patent-pending Dynamic Context Orchestration Algorithm (DCOA), MK3 generated its projection on September 12—well before nominations were filed and long before election momentum had taken shape.
According to MK3, the National Democratic Alliance (NDA) was expected to secure 138+ seats, while the Mahagathbandhan was projected to obtain 78+ seats. A comparative analysis conducted on November 14, based on Election Commission data and mainstream media reports, indicates that MK3 achieved a directional accuracy of 81.28%. This performance underscores the growing capability of Indian-developed AI systems to interpret complex public datasets and provide early, reliable insights.
Related Posts:
At the time of MK3’s prediction, the electoral landscape was still fluid. Candidates had yet to file nominations and major political announcements were still weeks away. Despite the absence of formal campaigning and constituency-level polling, MK3’s projection relied solely on historical, demographic, administrative and behavioral data, processed through its multilingual, context-aware architecture. The system was able to detect patterns and directional shifts earlier than traditional forecasting methods.
When counting began on November 14, trends closely mirrored MK3’s September projection, with the NDA leading in the 190–200+ seat range and the Mahagathbandhan significantly below earlier expectations.
“The close alignment between MK3’s September prediction and the actual trends demonstrates that Indian-developed AI systems can deliver meaningful insights even in complex scenarios,” said Gautom Das creator of MK3. He emphasised that the system was built to handle India’s linguistic and data diversity, highlighting the platform’s alignment with the Make in India and Digital India initiatives.
MK3’s accuracy score was calculated using a proportional error formula based on Bihar’s 243-seat assembly. The differences between predicted and actual seat counts for both alliances were measured, and the mean error produced the final figure of 81.28%. Developers noted that this methodology provides transparency and is suitable for multi-party electoral systems.
MK3, developed by MountKan R&D is a browser-based generative AI platform capable of operating in over 100 languages. Its DCOA technology allows it to interpret context over long sessions and adapt its reasoning or content generation according to user intent. Since its introduction in May 2025, MK3 has been positioned as an innovation aimed at improving accessibility, digital literacy and language inclusion for students, professionals and communities across India.
