Bengaluru startup Sarvam AI unveiled its Sarvam-30B and Sarvam-105B models at the India AI Impact Summit 2026 in New Delhi. Built by a team of about 15 young engineers, the 105B mixture-of-experts model reportedly outperforms leading systems on several benchmarks while supporting 22 Indian languages, highlighting India’s growing capability to develop frontier AI models.
When Bengaluru-based Sarvam AI unveiled its large language models at the India AI Impact Summit 2026 in New Delhi, the response was electric. The startup's pavilion drew record crowds at Bharat Mandapam, with thousands of students, developers and professionals queuing to witness live demonstrations - and organisers reportedly extended the expo by a day. The 105B model, built on a mixture-of-experts architecture with a 128,000-token context window, outperforms DeepSeek R1 and Google Gemini Flash on several benchmarks - a jaw-dropping achievement for a startup that barely existed three years ago.
Behind every benchmark and every demo are real people. Here are the engineers who gave blood, sweat, and code to make it happen. These engineers are making India proud.
1. Sumanth Doddapaneni: ML Researcher
A PhD student at IIT Madras currently on leave, Sumanth holds a BTech from IIIT SriCity and is a Google PhD Fellow (2023). His research on multilingual language modelling has earned Outstanding Paper Awards at both ACL 2024 and EMNLP 2024. He previously interned at Google Research in Bengaluru and Mountain View, and at Mila — Quebec AI Institute.
-
Ange Postecoglou's two-word response on Tottenham return with Igor Tudor on the brink
-
WAFCON 2026: A 'combination of factors' behind postponement
-
NCERT Unveils New English Textbook for Class 9 Featuring Prominent Indian Authors
-
BARC Junior Research Fellowship 2026: Application Details and Eligibility Criteria
-
Odisha SSB Junior Assistant and Clerk Recruitment 2026 Announced
