{"id":57626,"date":"2026-07-31T13:18:58","date_gmt":"2026-07-31T07:48:58","guid":{"rendered":"https:\/\/financialtelegraph.in\/index.php\/2026\/07\/31\/when-machines-learn-to-think-in-hindi-indias-bet-on-homegrown-ai\/"},"modified":"2026-07-31T13:18:58","modified_gmt":"2026-07-31T07:48:58","slug":"when-machines-learn-to-think-in-hindi-indias-bet-on-homegrown-ai","status":"publish","type":"post","link":"https:\/\/financialtelegraph.in\/index.php\/2026\/07\/31\/when-machines-learn-to-think-in-hindi-indias-bet-on-homegrown-ai\/","title":{"rendered":"When Machines Learn to Think in Hindi: India\u2019s Bet on Homegrown AI"},"content":{"rendered":"
\n

\"\"<\/p>\n

Mumbai (Maharashtra) [India], July 31:<\/strong> If you ask ChatGPT a question in proper Hindi, it\u2019ll usually handle it. But toss it the way people really talk\u2014mixing English and Hindi, typing in Roman letters to save time\u2014suddenly, it starts tripping up. Same goes for a farmer in Gujarat using Gujarati, or someone in rural Tamil Nadu using Tamil. The model gets English first; everything else is almost an afterthought. That\u2019s the gap a bunch of Indian AI startups are trying hard to close.<\/p>\n

The problem nobody in Silicon Valley was solving<\/h4>\n

Silicon Valley never really bothered with this problem. Big AI labs go after languages with mountains of internet text\u2014English, Chinese, Spanish. They don\u2019t touch Hindi, Tamil, Bengali, Telugu, or the eighteen other official Indian tongues, because there\u2019s not enough \u201cclean\u201d digital content for training. Sure, plenty of documents exist\u2014old books, newspapers, government reports\u2014but they\u2019re scattered all over and not easily fit for feeding into a big language model. And come on: making a model that actually gets the grammar and culture behind 22 languages at once? That\u2019s a way bigger headache than tweaking an English model and hoping for the best.<\/p>\n

This is exactly what Bengaluru-based Sarvam AI is tackling. At the India AI Impact Summit in New Delhi, February 2026, they launched Sarvam-30B and Sarvam-105B. Both models were built from scratch, instead of just reworking foreign ones. The big one, Sarvam-105B, uses a mixture-of-experts design, only lighting up a chunk of its brain at a time\u2014which means it\u2019s snappy enough for real conversations, even on basic devices. During the demo, the chatbot called itself \u201cVikram\u201d (a shout-out to Vikram Sarabhai, the guy who kickstarted India\u2019s space program) and casually talked in Hindi, Punjabi, and Marathi\u2014even on old-school feature phones.<\/p>\n

Why code-mixing is the real test<\/h4>\n

But what really makes these models shine isn\u2019t just their size. It\u2019s what they were trained on. Sarvam purposely blended formal text with the chaos of real life\u2014think Hinglish on WhatsApp, Telugu tossed into an English message, customer service chats flipping between languages. Most big models trip over this because they\u2019re used to polished, single-language training data. If your model can\u2019t keep up with a sentence bouncing between three languages before lunch, it\u2019s pretty much useless for most Indians.<\/p>\n

By March 2026, both models went open source and were showing their stuff. Sarvam-30B powers Samvaad, a platform for chatbots, while Sarvam-105B handles Indus, which is geared toward chunkier reasoning. Tech Mahindra rolled out Indus 2.0, blending NVIDIA\u2019s Nemotron-Hindi model with their own systems so businesses could cover Hindi dialects without blowing money on giant servers.<\/p>\n

Government money, government use cases<\/h4>\n

None of this happens by accident, either. The IndiaAI Mission has been quietly footing the bill for the heavy lifting, and their own big projects\u2014things like 2047: Citizen Connect and AI4Pragati\u2014stick to the same idea: let people talk to government services in the language they actually use, not whatever\u2019s written on official forms.<\/p>\n

India\u2019s wrestled with this for years. Look at Bhashini\u2014it\u2019s the government\u2019s ongoing work to bridge the language gap so that the 800 million Indians who aren\u2019t cool with English can actually use digital services. Sarvam\u2019s gamble is simple: you shouldn\u2019t have to translate your own question into English just to figure out a subsidy or an insurance claim.<\/p>\n

The commercial case is just as strong<\/h4>\n

Of course, the dollars make sense, too. Telecoms, banks, and e-commerce firms serve hundreds of millions in a dozen-plus languages, so far cobbling together rule-based bots, English help lines, and overloaded call centers. If a model knows Tamil, Telugu, and Hinglish without needing to convert everything into English first, support becomes way cheaper and smoother. That\u2019s why e-vikrAI\u2014an earlier Indian vision-language model\u2014helped sellers automate product listings for e-commerce, instead of forcing them to write everything out in three languages.<\/p>\n

Sovereignty, not just convenience<\/h4>\n

But honestly, there\u2019s another layer here: control. Building foundation models from scratch, using Indian servers, means the country isn\u2019t stuck renting its digital future from San Francisco. It\u2019s the same spirit that pushed UPI to crush global payment networks. Maybe Sarvam\u2019s models can\u2019t beat OpenAI or Google on pure IQ yet\u2014the tech benchmarks are still tough even for the big 105-billion model. But that\u2019s not really the point. The mission is to create something that understands a grandmother in Kanpur asking questions in the Hindi she actually uses, not proper textbook Hindi.<\/p>\n

That\u2019s a much narrower goal than \u201cbuild world\u2019s smartest AI,\u201d but honestly, it sounds a lot more useful\u2014especially when you\u2019ve got about 1.4 billion people depending on it.<\/p>\n

PNN Technology<\/a><\/strong><\/p>\n

\n<\/div>\n","protected":false},"excerpt":{"rendered":"

Mumbai (Maharashtra) [India], July 31: If you ask ChatGPT a question in proper Hindi, it\u2019ll usually handle it. But toss it the way people really talk\u2014mixing English and Hindi, typing … Read More<\/a><\/p>\n","protected":false},"author":1,"featured_media":57627,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[31],"tags":[670],"class_list":["post-57626","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-technology","tag-technology","entry"],"_links":{"self":[{"href":"https:\/\/financialtelegraph.in\/index.php\/wp-json\/wp\/v2\/posts\/57626","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/financialtelegraph.in\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/financialtelegraph.in\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/financialtelegraph.in\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/financialtelegraph.in\/index.php\/wp-json\/wp\/v2\/comments?post=57626"}],"version-history":[{"count":0,"href":"https:\/\/financialtelegraph.in\/index.php\/wp-json\/wp\/v2\/posts\/57626\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/financialtelegraph.in\/index.php\/wp-json\/wp\/v2\/media\/57627"}],"wp:attachment":[{"href":"https:\/\/financialtelegraph.in\/index.php\/wp-json\/wp\/v2\/media?parent=57626"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/financialtelegraph.in\/index.php\/wp-json\/wp\/v2\/categories?post=57626"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/financialtelegraph.in\/index.php\/wp-json\/wp\/v2\/tags?post=57626"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}