What is the common examiner trap about Science & Tech in UPSC exams?

They may confuse the parameter sizes (35B vs 105B) or swap the subsidy amount (₹100 crore) with the total cluster cost (₹246 crore). Another trap is attributing the launch to a different ministry or confusing the model architecture (MoE vs dense models).

What should UPSC aspirants remember about Science & Tech for Prelims?

1. Target this Data: Over 36,000 GPUs commissioned under IndiaAI Mission; Subsidy of almost ₹100 crore to Sarvam AI. 2. Target this Nodal Body: Ministry of Electronics and Information Technology (MeitY) - driving the IndiaAI Mission. 3. Target this Tech Term: Mixture of Experts (MoE) Architecture - key for cost-effective LLMs.

Science & TechnologyEconomy

Take AI Challenge ▸

News 28 of 30

IndiaAI Mission Subsidizes LLM Training with ₹100 Cr GPU Access, Sarvam AI Launches 35 & 105 Billion Parameter Models

Q: Which ministry is primarily responsible for the IndiaAI Mission that subsidizes LLM training in India?

Ministry of Electronics and Information Technology, The article explicitly states that the Ministry of Electronics and Information Technology (MeitY) has encouraged domestic LLM development for reasons including fostering the Indian AI ecosystem and addressing language representation issues.

Q: Consider the following statements regarding the news on Indian LLM training: 1. The IndiaAI Mission has commissioned over 36,000 GPUs in data centres operated by Indian firms like Yotta. 2. Sarvam AI's LLMs were trained on 35 billion and 105 billion parameters respectively and demonstrate improvements primarily in European languages. 3. A key reason for government support is the belief that foreign-developed LLMs cannot develop the capacity to work well with Indian languages. Which of the statements given above is/are correct?

Both 1 and 3, Statement 1 is correct as per the article. Statement 2 is incorrect because Sarvam AI's models demonstrated improvements over other models in Indian languages, not European languages. Statement 3 is correct as the article mentions this as a main reason for government encouragement.

Q: According to the article, what is the estimated government subsidy provided to Sarvam AI for accessing GPUs under the IndiaAI Mission?

Almost ₹100 crore, The article states: "The government gave Sarvam access to 4,096 GPUs from its common compute cluster, and the subsidy so far is estimated at almost ₹100 crore." The other figures are distractors; ₹246 crore is the 'bill of materials' for the cluster.

Q: What is a primary advantage of the Mixture of Experts (MoE) architecture for Large Language Models, as highlighted in the article?

It only activates a fraction of the overall parameters, making it run faster and consume fewer resources., The article explains that unlike earlier models which activated all parameters, an MoE model 'only activates a fraction of the overall parameters of a model, making it run faster and also consume a lower level of computing resources.' This is its key advantage for cost-effective deployment.

Science & Tech Economy PolityHIGH

Target:UPSC GS-IIIMPSCSSC GATeachingPrelims HighMains MediumStatic GK Link

26 Feb 2026

~2 min

Source: The Hindu

Key Data:35 billion parameters105 billion parametersOver 36,000 GPUs4,096 GPUs for Sarvam₹100 crore subsidy₹246 crore cluster cost

Bodies:Ministry of Electronics and Information TechnologySarvam AIBharatGenGnani.aiYotta

Practice MCQs from today's news ▸

~2 min · Quick mode

What This Article Covers

1.Bengaluru-based Sarvam AI launched two LLMs (35B & 105B parameters) trained with government-subsidized GPU access, focusing on Indian language performance.

2.The IndiaAI Mission subsidized training by commissioning over 36,000 GPUs, giving Sarvam access to 4,096 GPUs with an estimated subsidy of ₹100 crore.

3.The Mixture of Experts (MoE) architecture is highlighted as a key innovation for making LLMs faster and less resource-intensive, crucial for affordable local deployment.

The Big Picture

Prelims · HighMains · Medium

India is strategically subsidizing domestic AI development through the IndiaAI Mission, providing startups like Sarvam AI access to over 36,000 GPUs to train Large Language Models (LLMs) focused on Indian languages. This move aims to reduce reliance on foreign AI models and build a self-reliant, cost-effective AI ecosystem.

Exam Lens

Quick Exam Facts From News

Sarvam AI Models35 billion and 105 billion parameters

IndiaAI Mission GPU AccessOver 36,000 GPUs commissioned

Sarvam's GPU Access4,096 GPUs from common compute cluster

Estimated SubsidyAlmost ₹100 crore

Cluster 'Bill of Materials'₹246 crore

1-Minute Revision

›Sarvam AI Models: 35 billion and 105 billion parameters
›IndiaAI Mission GPU Access: Over 36,000 GPUs commissioned
›Target this Data: Over 36,000 GPUs commissioned under IndiaAI Mission; Subsidy of almost ₹100 crore to Sarvam AI.
›Target this Nodal Body: Ministry of Electronics and Information Technology (MeitY) - driving the IndiaAI Mission.
›Target this Tech Term: Mixture of Experts (MoE) Architecture - key for cost-effective LLMs.

Mastered this topic? Test your knowledge with a full MCQ quiz.

Practice exam-style questions, track your score, and strengthen your recall.

News Snapshot

🤖

Domestic AI Push

The IndiaAI Mission subsidizes LLM training to foster a domestic AI ecosystem and reduce reliance on foreign models for Indian languages.

💰

Subsidy & Infrastructure

A ₹100 crore subsidy provides startups access to over 36,000 GPUs, addressing the high capital and compute cost challenge for training LLMs.

🧠

MoE Architecture Advantage

Mixture of Experts (MoE) architecture makes LLMs cheaper and faster by activating only a fraction of parameters during inference.

Detailed Analysis & Strategic Depth

Context / Background

Global LLMs like ChatGPT are trained primarily on English and other widely-represented languages, leading to poor performance on Indian languages. Training LLMs requires massive GPU clusters and millions of dollars, creating a barrier for Indian firms. The government believes foreign-developed LLMs lack the incentive to optimize for Indian languages, necessitating a domestic push.

Core Analysis & Implications

The IndiaAI Mission's subsidy of ₹100 crore for GPU access tackles the capital scarcity problem head-on. Sarvam AI's models, though smaller than global 'frontier models', are optimized for the Indian context and efficiency using the MoE architecture. This strategy mirrors China's DeepSeek R1 model, which sparked industry-wide adoption due to cost advantages. The focus on Indian language performance addresses a critical gap, but the models' open-source availability for scrutiny remains limited as they are not on platforms like Hugging Face.

Future Outlook / Way Forward

The success of this initiative hinges on sustained investment for larger training runs to match the depth of paid global models. Wider adoption by developers and integration into sectors like education and healthcare (as seen with BharatGen's model) will be key metrics. The government aims to spark a cost-advantage ecosystem similar to China's, making affordable, localized AI a competitive national asset.

Key Facts at a Glance

Sarvam AI Models35 billion and 105 billion parameters

IndiaAI Mission GPU AccessOver 36,000 GPUs commissioned

Sarvam's GPU Access4,096 GPUs from common compute cluster

Estimated SubsidyAlmost ₹100 crore

Cluster 'Bill of Materials'₹246 crore

You've read the full analysis. Now test your understanding.

Practice exam-style questions based on this article.

Q1Static LinkageEasyUPSC, SSC, State PSC

Which ministry is primarily responsible for the IndiaAI Mission that subsidizes LLM training in India?

Q2Statement-basedHardUPSC, SSC

Consider the following statements regarding the news on Indian LLM training:

1. The IndiaAI Mission has commissioned over 36,000 GPUs in data centres operated by Indian firms like Yotta.

2. Sarvam AI's LLMs were trained on 35 billion and 105 billion parameters respectively and demonstrate improvements primarily in European languages.

3. A key reason for government support is the belief that foreign-developed LLMs cannot develop the capacity to work well with Indian languages.

Which of the statements given above is/are correct?

Q3Data-centricMediumUPSC, Banking, SSC

According to the article, what is the estimated government subsidy provided to Sarvam AI for accessing GPUs under the IndiaAI Mission?

Q4Application/ImpactMediumUPSC, SSC

What is a primary advantage of the Mixture of Experts (MoE) architecture for Large Language Models, as highlighted in the article?

All 25 MCQs ▸

Read original article from The Hindu

You finished this topic

Practice 3 MCQs from this news Next Article

Explore Related Topics

Science & TechView Topic MCQs →EconomyView Topic MCQs →PolityView Topic MCQs →

PreviousUS Imposes 126% Countervailing Duty on Indian Solar Cells Citing Adani Non-Cooperation & Chinese Inputs Next Karnataka HC Issues Guidelines for Spouses to Access Income Tax Returns in 14-Day Hearing for Maintenance Cases

Related Current Affairs

Science & Tech Current Affairs

Centre Launches Semicon 2.0: Rs 1.27 Lakh Cr Programme for Chip Design, Fabless Firms & Deployment Incentives

Govt launches Semicon 2.0 with Rs 1.27 lakh crore to boost indigenous chip design, support fabless companies, and provide deployment-linked incentives. The programme aims to build Indian-owned IP and reduce dependence on foreign semiconductors, targeting talent development and massive investment.

Science & Tech Current Affairs

ISRO Clarifies Pvt Sector Role: Targets $44B Space Economy, Bharatiya Antariksh Station by 2035, Moon Mission by 2040

ISRO has addressed employee concerns, stating that private sector participation in space is not a threat but an expansion of the ecosystem. The agency will focus on advanced missions like Bharatiya Antariksh Station (2035) and crewed Moon mission (2040), while routine production shifts to industry. India aims to grow its space economy from $8.4B to $44B by 2033.

Economy Current Affairs

₹62,500 Cr Smartphone Manufacturing Scheme: Two-Track Strategy for PLI Successor & Indian Brands

India's new Rs 62,500 crore smartphone scheme aims to sustain large-scale production after the first PLI ends and create a homegrown brand with its own IP. It offers two tracks: one for global manufacturers with incremental production targets, and another with higher incentives for Indian-owned companies that design and develop in India.

Science & Tech Current Affairs

ISRO to Focus on Big Missions, Private Sector to Handle Launches: SSLV Tech Transferred to HAL

ISRO is being repositioned like NASA – focusing on exploration and R&D while private industry takes over routine launches. IN-SPACe chairman's remark that ISRO will eventually stop making launch vehicles sparked debate, but ISRO chairman clarified no privatisation. The transfer of SSLV technology to HAL signals the new direction, but concerns over revenue loss and institutional independence remain.

Science & Tech Current Affairs

India's R&D Spending, Patents Surge but Systemic Barriers Hinder Science Superpower Status: E-book

A Hindu e-book analyzes India's scientific trajectory, noting increased R&D spending and patent filings but flagging critical gaps: poor lab-to-market translation, bureaucratic hurdles, brain drain, and rising pseudoscience. For exams, this is a rich source of 'Challenges to India's Innovation Ecosystem' for GS3.

Science & Tech Current Affairs

$10 Bn AI Investment Can Bring India to AI Frontier in 1 Year: Sarvam AI CEO

India's AI gap with the global frontier is just $10 billion, and with targeted investment, the country can catch up within a year, according to Sarvam AI CEO Pratyush Kumar. The India AI Impact Summit in February saw over $400 billion in AI capex commitments, and voice AI is emerging as India's first billion-dollar revenue opportunity. However, job creation and upskilling remain critical challenges.

Science & Tech Current Affairs

PM Modi Meets 20 Space Startups on National Space Day, Pledges Unwavering Support for Sector

PM Modi interacted with 20 space startups ahead of National Space Day, emphasizing India's emergence as a global space hub. He assured policy continuity and unwavering support as the sector sees a boom post-2020 liberalization, with Skyroot Aerospace becoming the first private Indian company to conduct an orbital launch.

Science & Tech Current Affairs

PM Modi to 20+ Space Start-ups: India's Private Space Ecosystem Can Attract Global Talent Pool

PM Modi met over 20 space start-up founders ahead of National Space Day, lauding their courage and stating that India's private space ecosystem can become a global magnet for scientific talent. The government promised stable policy support, boosting the morale of India's rapidly expanding space start-up sector, which now has over 400 companies.

Start Today's Quiz(25 Qs)

Take Today's Quiz

25 Qs • 5-7 Mins

IndiaAI Mission Subsidizes LLM Training with ₹100 Cr GPU Access, Sarvam AI Launches 35 & 105 Billion Parameter Models

Exam Lens

Quick Exam Facts From News

1-Minute Revision

News Snapshot

Domestic AI Push

Subsidy & Infrastructure

MoE Architecture Advantage

Detailed Analysis & Strategic Depth

Context / Background

Core Analysis & Implications

Future Outlook / Way Forward

Prelims Flashpoints

Key Facts at a Glance