Centre selects start-up Sarvam to build country's first homegrown AI model
The government has selected the Bengaluru-based start-up Sarvam to build the country's first indigenous artificial intelligence (AI) large language model (LLM) amid waves made by China's low cost model DeepSeek. The start-up, chosen from among 67 applicants, will receive support from the government in terms of compute resources to build the model from scratch.
Sarvam is the first start-up to get approved for sops under India's ambitious Rs 10,370 crore IndiaAI Mission to build a model, with the government currently assessing hundreds of other proposals. Sarvam said that its model will be capable of reasoning, designed for voice, and fluent in Indian languages, and it will be ready for population-scale deployment.
A senior official said that in terms of government support, the company will receive access to 4,000 graphics processing units (GPUs) for six months for the company to build and train its model. The model is not expected to be open-sourced, but will be fine tuned particularly for Indian languages. The GPUs will be provided to Sarvam by the companies which have been separately selected by the government to set up AI data centres in India.
'This (Sarvam's) model will have 70 billion parameters and many innovations in programming as well as engineering. With these innovations, a 70 billion parameter (model) can compete with some of the best in the world,' said IT Minister Ashwini Vaishnaw.
As part of the Sarvam's LLM proposal, the company is developing three model variants: Sarvam-Large for advanced reasoning and generation, Sarvam-Small for real-time interactive applications, and Sarvam-Edge for compact on-device tasks, said Pratyush Kumar, one of the the company's two co-founders.
The development comes amid the meteoric rise of DeepSeek, a low-cost foundational model from China, which shook up the AI industry. DeepSeek's entry into the AI space – touted for being open source, its accuracy and claims that it has been built at a fraction of the cost as its US competitors – sent Nvidia's stock on a downward spiral, since its R1 model was trained on inferior GPUs compared with the likes of OpenAI.
Sarvam's model will be built, deployed, and optimised in India, using local infrastructure and developed by a new generation of Indian talent. This initiative aims to promote strategic autonomy, accelerate domestic innovation, and secure India's leadership in AI for the long term, the company said in a press release.
Vivek Raghavan, co-founder of Sarvam, said, 'This is a crucial step toward building critical national AI infrastructure. Our goal is to build multi-modal, multi-scale foundation models from scratch. When we do, a universe of applications unfolds. For citizens, this means interacting with AI that feels familiar, not foreign. For enterprises, this means unlocking intelligence without sending their data beyond borders'.
Earlier this year, the government had also selected 10 companies to supply 18,693 GPUs — high-end chips needed to develop machine learning tools — that can go into developing a foundational model. This is more than the initial aim of the IndiaAI Mission, under which the government was looking to procure 10,000 GPUs.
The companies empaneled to provide the GPU services include Jio Platforms, the Hiranandani Group-backed Yotta, Tata Communications, E2E Networks, NxtGen Datacenter, CMS Computers, Ctrls Datacenters, Locuz Enterprise Solutions, Orient Technologies, and Vensysco Technologies.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Time of India
33 minutes ago
- Time of India
MU revokes paid maternity leave for guest faculty and temporary staff
Mangaluru: In a cost-cutting move, Mangalore University (MU) has revoked paid maternity leave for guest faculty and temporary non-teaching staff. The syndicate granted four months of paid leave in July 2022, but the benefit was discontinued last month without alternative provisions. Tired of too many ads? go ad free now The cash-strapped MU aims for this move to significantly improve its financial condition, and the syndicate pointed out that there was no necessity to provide such benefits to guest faculty and temporary staff. Judith Mendonca, a syndicate member who proposed scrapping the policy, stated that guest faculty working under MU are paid a substantial amount of compensation compared to other state universities in Karnataka. The monthly salary is around Rs 40,000. Despite paying such a huge amount, their workload is minimal. There are guest faculty appointed for courses with fewer than 10 students. They have a maximum of three hours of work per day. In this scenario, it is an absolute financial burden on MU to offer the temporary staff such benefits. Therefore, MU has decided to withdraw and scrap the rule from May 2025 with immediate effect," explained Mendonca. The decision to offer paid maternity leave was taken after former syndicate member Ramesh K tabled the proposal. He informed that one of the guest faculty working in Kodagu district, which was earlier under MU, requested that all women guest lecturers be considered for paid maternity leave. No guest faculty or temporary staff under MU were eligible for the facility, even though many worked for more than a decade. The aim was to give them relief as they have no pay until they return to work. Tired of too many ads? go ad free now Currently, only contractual employees are eligible for six months of paid maternity leave, as directed by the state govt. Unhappy with the decision of MU, female guest faculty members said that they need to be given some other benefits for their welfare or reverse the decision.
Indian Express
41 minutes ago
- Indian Express
What's behind traffic snarls at 5 key border points in Delhi?
Daily traffic chaos along Delhi-NCR borders is more than just routine rush hour chaos — it's also fueled by toll booths and plazas run by the Municipal Corporation of Delhi (MCD). According to a recent government report, these toll points are the primary culprits behind the massive jams seen during morning and evening peak hours. Five key Delhi borders, as per the document, face daily traffic due to the MCD toll booths: Sarhaul/Rajokri border on NH-48; Delhi-Meerut Expressway; Dwarka Expressway near Bijwasan; Badarpur border (MCD plaza on NH-19/44); Urban Extension Road II (UER II) at Jharoda Kalan (near Najafgarh-Bahadurgarh Road). The issue was raised at a recent meeting between the Centre and the Delhi government on a plan to decongest the city, which includes ongoing big infrastructure projects being executed by the National Highways Authority of India (NHAI) to improve connectivity between Delhi and neighbouring cities. There are over 150 MCD toll booths, of which 13 are RFID (Radio-Frequency Identification) enabled. Over 1 lakh commercial vehicles enter Delhi daily, of which more than 50% are cabs. The civic body earns around Rs 800 crore a year from toll charges. At the meeting, officials said, the NHAI highlighted that significant traffic congestion at Delhi's borders is attributed to the MCD toll plaza operations. 'Traffic jams are witnessed due to the heavy flow of commercial vehicles at the specified locations. It has also been observed that traffic speed decreases at these [five] locations, raising the risk of collisions with high-speed vehicles approaching from behind…,' said officials. Officials also highlighted that the MCD has deployed a fee-collecting agency at these locations, causing hindrance to the free flow of traffic during peak hours. The MCD, in its response to the issues highlighted by the NHAI, said, 'Traffic snarls are taking place due to narrowing/funnelling of road just before the boundary of Airport and merging of UER-II traffic…' Officials said a possible solution to address this issue is implementing the gantry-based Multi-Lane-Free-Flow (MLFF) tolling system with FASTag integration. The MLFF system has no physical plaza. The tolling system will be mounted on gantries, which will capture vehicle information as they pass through. 'The MLFF system enables seamless, real-time toll collection using passive RFID and ANPR (Automatic Number Plate Recognition) technologies. As vehicles approach the gantry, their FASTag and license plates are automatically scanned. This allows the system to electronically deduct the appropriate toll amount from the driver's prepaid account without the need to stop or slow down,' said a senior official. Officials said that as the NHAI is the expert in the field of tolling, the MCD is planning to seek support from it. Officials from the corporation will visit and study NHAI's successful toll sites. 'Apart from this, the MCD is also planning to protect its revenue by interlinking FASTag with bank accounts, permits, insurance, etc,' said officials. In the meantime, the NHAI has asked the civic body to expedite the process of setting up the MLFF tolling systems at the five border areas. 'The Indian Highways Management Company Limited (IHMCL) may be contacted for providing necessary support to MCD for fastag/tolling integration,' read the document. The meeting also discussed other issues, including a request from the NHAI to the MCD to deposit toll tax collections that were due after March 22, 2020. According to the document, an MoU was signed on July 12, 2019, between NHAI and the erstwhile South Delhi Municipal Corporation for collecting toll at the Badarpur Elevated Fee Plaza on NH-2. The agreement was valid from June 12, 2019, to July 12, 2020. However, SDMC has not deposited the agreed amount since March 22, 2020. NHAI has requested a payment of Rs 21.3 crore, which has now increased to Rs 23.3 crore due to interest. In response, the MCD said it is trying to recover the amount from the contractor who was handling toll collection at the time. Officials added that several cases have been filed in different courts, and the matter is currently sub-judice.
Hindustan Times
42 minutes ago
- Hindustan Times
India seeks exemption from US' 10% baseline tariff
The fate of the 10% baseline tariffs that the Trump administration invoked on imports from all countries on April 2 is among the issues now at the heart of negotiations between New Delhi and Washington as they attempt to hammer out an early tranche of the trade deal, people aware of the matter have said. Delhi is not in favour of replicating, as suggested by the American negotiators, the approach in the trade deal struck between the US and the UK, where British goods are still subject to the baseline tariffs, these people added. According to a person with direct knowledge of the discussions, Indian negotiators are pushing for their American counterparts to remove the baseline 10% rate as well as commit to assurances that the additional 16%, due to be implemented on July 9, will be left off. An American negotiating team led by assistant US Trade Representative Brendan Lynch 4 landed in Delhi on June for what is the fifth time negotiators from either side have gone to the other's capital for face-to-face talks. The American delegation is expected to be in Delhi till June 10, longer than the previously expected two-day visit. ALSO READ | India-US trade negotiations hit top gear, American delegation extends Delhi stay 'Ideally, both the 10% baseline tariff on Indian goods and the additional 16% from July 9 must end simultaneously after an interim deal is signed. Else, India will also have rights to continue proportionately similar tariffs on American goods till the time the US withdraws the entire 26% reciprocal tariff,' one of them said, citing a joint statement by the two countries' leaders issued on February 13 in Washington. While expounding 'Mission-500' to double bilateral trade to $500 billion by 2030 on February 13, the two leaders – Prime Minister Narendra Modi and President Donald Trump – in their joint statement mentioned the need for new 'fair-trade terms' that are 'mutually beneficial', the person said. A second person aware of the matter corroborated India's stance: 'Only a mutually beneficial deal would have a long life'. 'Both India and the US are sovereigns. One is the oldest democracy and the other is the largest democracy. While the US is the largest economy, India is the fastest growing major economy of the world. Hence, the deal must be balanced, equitable, fair and acceptable to their people,' the first person said. The second person added that India sees trade interests between both nations as being 'complementary and not competitive', hence New Delhi is open to giving greater market access to the American goods in the Indian market provided Washington reciprocates. 'The trade negotiations continue in New Delhi covering all these matters in a constructive manner as we speak and both sides are hopeful for a win-win,' he said. ALSO READ | Donald Trump claims India willing to cut 100% tariffs on US goods, 'but…' After UK industries faced American tariffs of 25% on all aluminium, steel and derivatives (announced on March 12), 25% tariff on passenger vehicles (announced on April 3), 25% tariff on automobile parts (beginning May 3), and a 10% baseline tariff on all imports (from April 5) – the UK and the US on May 8 announced an economic prosperity deal (EPD). The mini deal secured some concessions for the UK, but the 10% baseline tariff continued. Both partners are racing to conclude an interim, or regarded as an 'early harvest', deal before July 9, which will be followed by a wider first tranche of Bilateral Trade Agreement (BTA) by September-October 2025. After that a comprehensive BTA will be negotiated, they said. ALSO READ | How Donald Trump decided the tariff for India The current negotiations for an early harvest deal involve greater market access for goods by eliminating tariffs and non-tariff trade barriers, and improving supply chain integration, they said. The current New Delhi round is followed by a face-to-face negotiation between the two teams in the US. During that period, Union commerce and industry minister Piyush Goyal was also in the US from May 17-22 where he held meetings with his counterparts, US commerce secretary Howard Lutnick and USTR Jamieson Greer.
