Dark | Light
# ![@pratykumar Avatar](https://lunarcrush.com/gi/w:26/cr:twitter::1530518342049890307.png) @pratykumar Pratyush Kumar

Pratyush Kumar posts on X about drop, ai, we are, india the most. They currently have [------] followers and [--] posts still getting attention that total [-------] engagements in the last [--] hours.

### Engagements: [-------] [#](/creator/twitter::1530518342049890307/interactions)
![Engagements Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1530518342049890307/c:line/m:interactions.svg)

- [--] Week [---------] +569%
- [--] Months [---------] +1,421%
- [--] Year [---------] +3,192%

### Mentions: [--] [#](/creator/twitter::1530518342049890307/posts_active)
![Mentions Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1530518342049890307/c:line/m:posts_active.svg)

- [--] Year [--] no change

### Followers: [------] [#](/creator/twitter::1530518342049890307/followers)
![Followers Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1530518342049890307/c:line/m:followers.svg)

- [--] Week [------] +50%
- [--] Months [------] +98%
- [--] Year [------] +440%

### CreatorRank: [------] [#](/creator/twitter::1530518342049890307/influencer_rank)
![CreatorRank Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1530518342049890307/c:line/m:influencer_rank.svg)

### Social Influence

**Social category influence**
[countries](/list/countries)  11.36% [finance](/list/finance)  4.55% [products](/list/products)  2.27%

**Social topic influence**
[drop](/topic/drop) #171, [ai](/topic/ai) #1317, [we are](/topic/we-are) #304, [india](/topic/india) #1314, [build](/topic/build) #3499, [data](/topic/data) #1250, [agentic](/topic/agentic) #9, [language](/topic/language) #523, [realtime](/topic/realtime) #181, [streaming](/topic/streaming) #1786

**Top accounts mentioned or mentioned by**
[@sarvamai](/creator/undefined) [@paramendra](/creator/undefined) [@officialkrishd](/creator/undefined) [@aiexplorations](/creator/undefined) [@kingofdwolrd](/creator/undefined) [@siddarthpaim](/creator/undefined) [@j_bindra](/creator/undefined) [@adarshatva1](/creator/undefined) [@mrityuandprem](/creator/undefined) [@alphonso269536](/creator/undefined) [@pk00202500](/creator/undefined) [@belliethomas](/creator/undefined) [@insignifictcman](/creator/undefined) [@generalusername](/creator/undefined) [@republic](/creator/undefined) [@thekrishnambiar](/creator/undefined) [@thedesignobsess](/creator/undefined) [@aravsrinivas](/creator/undefined) [@uidai](/creator/undefined) [@huggingface](/creator/undefined)
### Top Social Posts
Top posts by engagements in the last [--] hours

"Beyond just transcription Saaras V3 excels in three areas. It can automatically detect the language of the audio. It now comes with word-level time-stamps. And it also supports diarisation for multi-speaker audios. Try out your polyglot skills in the blog"  
[X Link](https://x.com/pratykumar/status/2021604299382345886)  2026-02-11T15:16Z 14.4K followers, 12.4K engagements


"@aiexplorations Agree - there is much improvement to be done. Both in the schwa representation and in intonation. Happy to collaborate"  
[X Link](https://x.com/pratykumar/status/2020721748157604197)  2026-02-09T04:49Z 13.4K followers, [--] engagements


"In a head-to-head comparison against leading AI dubbing platforms experts preferred Sarvam Studios outputs as the outright choice. The dubbed voice retains the tone pacing and expressiveness of the original speaker. For fine adjustments such as audio-visual alignment content teams can edit and regenerate specific segments directly in the Studio. https://twitter.com/i/web/status/2022251125781774450 https://twitter.com/i/web/status/2022251125781774450"  
[X Link](https://x.com/pratykumar/status/2022251125781774450)  2026-02-13T10:06Z 14.4K followers, [----] engagements


"In a head-to-head comparison of real-world document translations Studio achieved the highest translation quality rating. Our outputs were recognised to have the most consistent quality across complex domains including academic fiction and legal content. Also Studio maintains the original structure and visual integrity such as preserving tables headings diagrams and page hierarchy without requiring manual redesign"  
[X Link](https://x.com/pratykumar/status/2022251134115832152)  2026-02-13T10:06Z 14.4K followers, [----] engagements


"Drop 1/14: We pulled off a first of its kind. The budget speech by the Finance Minister was live telecast in multiple languages on @republic. Reached millions of homes with a latency under [--] minutes. https://www.youtube.com/watchv=wiQ9KesTBbA https://www.youtube.com/watchv=wiQ9KesTBbA"  
[X Link](https://x.com/pratykumar/status/2018027623973278107)  2026-02-01T18:24Z 14.2K followers, 37.1K engagements


"Complex multi-column layouts of an old newspaper in Malayalam"  
[X Link](https://x.com/pratykumar/status/2019408684192002154)  2026-02-05T13:52Z 14.1K followers, 11.4K engagements


"Odisha is positioning itself to become Indias compute backbone leveraging its unique position today of being both power and revenue surplus. Our partnership envisages building a 50MW AI-optimised facility that will function as the States AI public utility - powering e-governance agriculture advisories healthcare triage unified citizen helplines AI tutors and disaster management. This hub will also attract innovative companies into the state by providing access to cutting-edge AI infra thereby transitioning from being a mine-driven to a mind-driven economy."  
[X Link](https://x.com/pratykumar/status/2020724185073619161)  2026-02-09T04:59Z 13.7K followers, 11.5K engagements


"First a little more elaboration on the results. We picked up a standard ETL task of extracting about [---] finance and business metrics from about [--] PDF documents pertaining to a companys performance across years. We want an agentic system to parse the documents and extract the metrics as per a given schema. Accuracy is measured as the percentage of the metrics correctly extracted. When we threw Claude Code at this problem we managed low accuracy values of about 17% at costs of about $12 for a run taking over [--] minutes. We then built a multi-agent system with Arya in six design iterations"  
[X Link](https://x.com/pratykumar/status/2021116217243881565)  2026-02-10T06:57Z 13.8K followers, [----] engagements


"Now lets say a little more about our design motivation. Arya starts from a simple observation - current orchestration frameworks fail on two fronts. They lack runtime reliability with no infrastructure-grade guarantees on execution semantics inter-agent communication or state management. And they lack rapid iterability with no formal design language step-wise replay or modular observability. To address this we built Arya with four principles: Composable primitives: Every system is composed of only one of eight primitives: LLM Agent MCP Node Ledger Task Graph Code Interpreter and Artefact."  
[X Link](https://x.com/pratykumar/status/2021116223011094719)  2026-02-10T06:57Z 13.7K followers, [----] engagements


"I go around saying we should be builders across the 'full stack' @SarvamAI - from compute to data to models to apps. And in come @thekrishnambiar and @thedesignobsess and say we should compose our own music for model launches And so here we go with our first Sarvam sound track. India's most ambitious decade to build is here; come join the build. And yes for your ringtone upgrade go here: https://www.sarvam.ai/bulbul-audio https://www.sarvam.ai/bulbul-audio"  
[X Link](https://x.com/pratykumar/status/2020176703151358151)  2026-02-07T16:43Z 14.4K followers, 56.5K engagements


"Drop 7/14: Introducing Sarvam Arya - our multi-agent orchestration platform built from the ground up with robust systems engineering principles and frontier AI-assisted developer experience. We illustrate the Arya advantage on a common ETL task of extracting structured data from unstructured documents. Arya system with GPT [---] mini achieved 5x higher accuracy at 10x lower cost compared to Claude Code with agent swarm. Read more about Arya in this thread and our blog: https://www.sarvam.ai/blogs/introducing-sarvam-arya https://www.sarvam.ai/blogs/introducing-sarvam-arya"  
[X Link](https://x.com/pratykumar/status/2021116213599010969)  2026-02-10T06:57Z 14.4K followers, 132.4K engagements


"On each of the most popular [--] languages on the IndicVoices benchmark Saaras V3 outperforms leading models such as Deepgram Nova-3 Elevenlabs Scribe v2 Gemini [--] Pro and GPT-4o Transcribe. The Realtime streaming model achieves much lower latency while being within a percentage point of the best error rate. https://twitter.com/i/web/status/2021604280583807054 https://twitter.com/i/web/status/2021604280583807054"  
[X Link](https://x.com/pratykumar/status/2021604280583807054)  2026-02-11T15:16Z 14.4K followers, 14.8K engagements


"From lecture videos to national addresses from textbooks to novels Sarvam Studio is powering creation of multilingual content. We are excited to see what you will build. Read our blog for more details and to get started. https://www.sarvam.ai/blogs/sarvam-studio https://www.sarvam.ai/blogs/sarvam-studio"  
[X Link](https://x.com/pratykumar/status/2022251138469577134)  2026-02-13T10:06Z 14.4K followers, [----] engagements


"Fully agree on the core necessity for affordable and efficient AI. We need to dramatically increase per capita access to intelligence without ballooning per capita energy costs. What Sarvam AI has demonstrated - world class AI can be done much more affordably and sustainably with much lower energy and compute footprint - is important. I have come to believe that in software code generation much more energy and compute efficient approaches to AI can be What Sarvam AI has demonstrated - world class AI can be done much more affordably and sustainably with much lower energy and compute footprint"  
[X Link](https://x.com/pratykumar/status/2022502949763846624)  2026-02-14T02:47Z 14.4K followers, 15.1K engagements


"As much about our TTS model as is about the impeccable diction that Sanskrit affords. Code x Culture. Create these beautiful cards here - https://dashboard.sarvam.ai/text-to-speech Stress testing Bulbul v3 https://t.co/2hw4MUuBz9 https://dashboard.sarvam.ai/text-to-speech Stress testing Bulbul v3 https://t.co/2hw4MUuBz9"  
[X Link](https://x.com/pratykumar/status/2020680980600930737)  2026-02-09T02:07Z 14.4K followers, 17.6K engagements


"Drop 9/14: Today we are introducing Sarvam Studio our product to help creators go multilingual. One piece of content every corner of India. With AI video dubbing Studio generates high-fidelity dubs in [--] Indian languages. In an expert study participants preferred Sarvam Studio for overall quality and production readiness. With agentic document translation Studio excels in contextually translating long-form content across genres. Our evaluations demonstrate that readers strongly preferred the output from Studio across different genres. https://twitter.com/i/web/status/2022251117359636843"  
[X Link](https://x.com/pratykumar/status/2022251117359636843)  2026-02-13T10:06Z 14.4K followers, 227.1K engagements


"Drop 10/14: Announcing Sarvam Edge our dedicated effort to bring intelligence to run offline and on-device. Our goal is to make AI that is efficient private and accessible everywhere. Today we are showcasing our speech recognition speech synthesis translation and document digitisation models running on-device. This comes on the back of our research efforts to make models super small in memory and compute footprints while being close to accuracy of much larger models. We believe running models at the edge will truly unlock population scale AI in India. And we are working with partners to make"  
[X Link](https://x.com/pratykumar/status/2022727760045838458)  2026-02-14T17:40Z 14.4K followers, 20.7K engagements


"Two weeks of daily drops of what we have been building at @SarvamAI starts today"  
[X Link](https://x.com/pratykumar/status/2017787705816715552)  2026-02-01T02:30Z 14.4K followers, 68.9K engagements


"Drop 2/14: Sarvam Audio: a state-space based efficient audio language model that defines the new benchmarks in speech recognition for Indian languages. Significantly outperforms Gemini [--] and GPT 4o Transcribe in a range of benchmarks. See details in our blog: @SarvamAI https://www.sarvam.ai/blogs/sarvam-audio https://www.sarvam.ai/blogs/sarvam-audio"  
[X Link](https://x.com/pratykumar/status/2018489867257409560)  2026-02-03T01:01Z 14.4K followers, 69.9K engagements


"Drop 4/14: Introducing Sarvam Vision: a state-space based [--] billion parameter vision language model that is competitive with the best results in digitisation in English and defines a significantly higher bar for Indian languages. See the details in our blog: https://www.sarvam.ai/blogs/Sarvam-vision https://www.sarvam.ai/blogs/Sarvam-vision"  
[X Link](https://x.com/pratykumar/status/2019408640915186155)  2026-02-05T13:51Z 14.4K followers, 585.8K engagements


"Sarvam Vision achieves state-of-the-art accuracy of 84.3% on the olmOCR-Bench (English only subset) outperforming frontier models like Gemini [--] Pro and recent OCR models like DeepSeek OCR 2"  
[X Link](https://x.com/pratykumar/status/2019408645981892976)  2026-02-05T13:51Z 14.4K followers, 94.9K engagements


"Older book with long form Tamil"  
[X Link](https://x.com/pratykumar/status/2019408678974296290)  2026-02-05T13:52Z 14.4K followers, 11.4K engagements


"Drop 5/14: Introducing Bulbul V3 our latest text-to-speech model. It raises the bar for how human it sounds while being super robust. In an independent third-party human listening study Bulbul V3 delivers the highest listener preference and low error rates across use-cases and languages. See details in our blog but first watch the video. https://www.sarvam.ai/blogs/bulbul-v3 https://www.sarvam.ai/blogs/bulbul-v3"  
[X Link](https://x.com/pratykumar/status/2020023442842743006)  2026-02-07T06:34Z 14.4K followers, 386.2K engagements


"Bulbul V3 is live. This is your month to go all in. Unlimited usage open through February. The mic is yours https://dashboard.sarvam.ai/text-to-speech https://dashboard.sarvam.ai/text-to-speech"  
[X Link](https://x.com/pratykumar/status/2020023458705649702)  2026-02-07T06:34Z 14.4K followers, 11.2K engagements


"Drop 6/14: @SarvamAI is proud to announce a landmark in Indias sovereign AI journey through strategic partnerships with the Governments of Odisha and Tamil Nadu. The aim of these partnerships is to drive transformation by building at-scale compute sovereign models and the institutional capacity to drive AI adoption. Instead of going bottom-up with PoCs which find it hard to take off these partnerships will go top down with a state-wide mandate to deploy AI as a public utility embedding intelligence across all departments"  
[X Link](https://x.com/pratykumar/status/2020724181160415673)  2026-02-09T04:59Z 14.4K followers, 54K engagements


"The advantages of setting up sovereign AI cannot be over-stated. Today we face a digital asymmetry. We are exporting data and importing tokens. When our data flows outward it trains foreign models and the learning loops compound elsewhere leaving us with tokens but no leverage. States that move quickly will build enduring advantage. A case in point is the U.S. wherein AI drove nearly 3% GDP growth in [----]. If states in India can generate even 1% annual growth via AI the compounding effect towards [----] can be massive"  
[X Link](https://x.com/pratykumar/status/2020724191465820562)  2026-02-09T04:59Z 14.4K followers, 27.5K engagements


"The other key insight that Arya is based on is the following. Today there are two ends of the spectrum in agentic workflows. Simple LLM chains that run autonomously but handle only straightforward tasks. And frontier coding agents that tackle complex problems but require heavy human interaction. The missing middle is production-grade agents - systems that are genuinely sophisticated but run at scale with reliability thousands of times a day without a human in the loop. For this missing middle of agents frontier models can be moved from runtime to compile time. Like compilers they can"  
[X Link](https://x.com/pratykumar/status/2021116226345501111)  2026-02-10T06:57Z 14.4K followers, [----] engagements


"We plan to open-source Arya as this layer should be shared infrastructure the way databases and container orchestration are. This release will include a fully containerised runtime that can run anywhere a debug/replay user-interface for reviewing runs and an LLM-powered developer experience for authoring and deployments. Code coming soon. Meanwhile we are building for our enterprise partners with Arya. https://www.sarvam.ai/blogs/introducing-sarvam-arya https://www.sarvam.ai/blogs/introducing-sarvam-arya"  
[X Link](https://x.com/pratykumar/status/2021116229357048174)  2026-02-10T06:57Z 14.4K followers, 14.9K engagements


"India's decade to build with the values we stand for The all-new Sarvam is here. Designed for all of India. https://t.co/O78H03Jcki https://t.co/zxtycbcDAu The all-new Sarvam is here. Designed for all of India. https://t.co/O78H03Jcki https://t.co/zxtycbcDAu"  
[X Link](https://x.com/pratykumar/status/2021434345383395439)  2026-02-11T04:01Z 14.4K followers, 54.3K engagements


"Drop 8/14: Introducing Saaras V3 the next iteration of our speech recognition model. We extend our lead in this space with an even more accurate model particularly for mixed-language and noisy speech. We have also expanded support for all the [--] scheduled languages of India. And we now support real-time streaming In this mode the model delivers low latency while preserving transcription quality. https://www.sarvam.ai/blogs/asr https://www.sarvam.ai/blogs/asr"  
[X Link](https://x.com/pratykumar/status/2021604272912376156)  2026-02-11T15:16Z 14.4K followers, 145.8K engagements


"The streaming mode is fun: you can literally use it with a movie playing in the background Try it out here: https://www.sarvam.ai/blogs/asr#asr-sarvam https://www.sarvam.ai/blogs/asr#asr-sarvam"  
[X Link](https://x.com/pratykumar/status/2021604293900411367)  2026-02-11T15:16Z 14.4K followers, [----] engagements


"Whether you are building voice bots subtitling long audios or analysing call transcripts you would find value in building with Saaras V3. Get started with our APIs today: https://dashboard.sarvam.ai/speech-to-text https://dashboard.sarvam.ai/speech-to-text"  
[X Link](https://x.com/pratykumar/status/2021604302842716576)  2026-02-11T15:16Z 14.4K followers, [----] engagements


"This video shows document intelligence on device. An Odia document is digitised entirely on a MacBook Pro. We believe a range of such models will come bundled in productivity focused laptops"  
[X Link](https://x.com/pratykumar/status/2022727765594894609)  2026-02-14T17:40Z 14.4K followers, [----] engagements


"In this second example we demonstrate a hybrid edge-cloud architecture applied to a stock broking app. The user can fetch live data query a portfolio execute trades and review research all with voice processing running locally and LLMs and APIs on the cloud. We believe apps will be able to provide convenient hands-free voice experiences at super low operating costs. https://twitter.com/i/web/status/2022727772632998131 https://twitter.com/i/web/status/2022727772632998131"  
[X Link](https://x.com/pratykumar/status/2022727772632998131)  2026-02-14T17:40Z 14.4K followers, [----] engagements


"In this demo we show real-time voice translation between Indian languages. The user can speak in their preferred language and the system translates and responds in another preserving both meaning and natural delivery. Speech recognition translation and expressive text-to-speech work together seamlessly"  
[X Link](https://x.com/pratykumar/status/2022730614890545347)  2026-02-14T17:52Z 14.4K followers, [----] engagements


"Sarvam Edge is being developed in close collaboration with leading global device manufacturers combining Sarvam's models with years of hardware R&D to power the next generation of consumer devices. If you are an app developer who would like to use our Edge SDK or a device manufacturer who would like to bundle our models then reach out at edge@sarvam.ai https://www.sarvam.ai/blogs/sarvam-edge https://www.sarvam.ai/blogs/sarvam-edge"  
[X Link](https://x.com/pratykumar/status/2022731042399117479)  2026-02-14T17:53Z 14.4K followers, [----] engagements


"@AravSrinivas Aravind at @SarvamAI we are building sovereign models that combine deep reasoning and Indic language skills. Would love to have you join this mission"  
[X Link](https://x.com/pratykumar/status/1882053988323147897)  2025-01-22T13:13Z [----] followers, 114.7K engagements


"Next week I am in the Bay Area looking to meet exceptional researchers in the AI space. Come see what we have been building at @SarvamAI and what you can do to innovate for a billion Indians RSVP with a DM"  
[X Link](https://x.com/pratykumar/status/1899882901493190761)  2025-03-12T17:59Z [----] followers, 22.2K engagements


"Delighted to be working with @UIDAI in bringing AI to one of the most crucial tech infrastructures of our country"  
[X Link](https://x.com/pratykumar/status/1902241228772475315)  2025-03-19T06:10Z [----] followers, 13.7K engagements


"India needs to be a builder and not just a consumer in this defining era of technology. The journey towards strategic autonomy starts now DMs open. It's official We're proud to announce that Sarvam has been selected by the Government of India under the IndiaAI Mission to build India's sovereign Large Language Model. Building India's sovereign model from the ground up is a crucial step toward Atmanirbhar Bharat. The https://t.co/nA4WgQ0aw0 It's official We're proud to announce that Sarvam has been selected by the Government of India under the IndiaAI Mission to build India's sovereign Large"  
[X Link](https://x.com/pratykumar/status/1916082271561428995)  2025-04-26T10:49Z [----] followers, 93.6K engagements


"Clarification: This is not a grant. A gov body will take equity in Sarvam for the compute we receive. And we are committed to building public interest use-cases and enabling the ecosystem in various ways such as hyper-optimising the inferencing costs in India. BTW we as a country have a long way to go in AI. This is the time to strengthen each other. Let's go"  
[X Link](https://x.com/pratykumar/status/1916368769414533292)  2025-04-27T05:48Z [----] followers, 63K engagements


"Great to be receiving feedback on Sarvam-M. Please keep them coming. Will help strengthen our pipelines as we start to train our sovereign model. This was particularly interesting - https://alokbishoyi.com/blogposts/sarvam-political-evaluation.html https://alokbishoyi.com/blogposts/sarvam-political-evaluation.html"  
[X Link](https://x.com/pratykumar/status/1926545350883496168)  2025-05-25T07:46Z [----] followers, 58.4K engagements


"Good to see folks building creatively with Sarvam-M which is now trending on @huggingface 🤗 More coming soon"  
[X Link](https://x.com/pratykumar/status/1927643125473771742)  2025-05-28T08:28Z [----] followers, 38.9K engagements


"Was a fun chat with @mukeshbansal06 on LLMs strategic autonomy what Indian businesses can do in AI and consciousness vs intelligence - https://youtu.be/89ZqoXfdGT4 https://youtu.be/89ZqoXfdGT4"  
[X Link](https://x.com/pratykumar/status/1932826084267139460)  2025-06-11T15:43Z [----] followers, 14.8K engagements

Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing

@pratykumar Avatar @pratykumar Pratyush Kumar

Pratyush Kumar posts on X about drop, ai, we are, india the most. They currently have [------] followers and [--] posts still getting attention that total [-------] engagements in the last [--] hours.

Engagements: [-------] #

Engagements Line Chart

  • [--] Week [---------] +569%
  • [--] Months [---------] +1,421%
  • [--] Year [---------] +3,192%

Mentions: [--] #

Mentions Line Chart

  • [--] Year [--] no change

Followers: [------] #

Followers Line Chart

  • [--] Week [------] +50%
  • [--] Months [------] +98%
  • [--] Year [------] +440%

CreatorRank: [------] #

CreatorRank Line Chart

Social Influence

Social category influence countries 11.36% finance 4.55% products 2.27%

Social topic influence drop #171, ai #1317, we are #304, india #1314, build #3499, data #1250, agentic #9, language #523, realtime #181, streaming #1786

Top accounts mentioned or mentioned by @sarvamai @paramendra @officialkrishd @aiexplorations @kingofdwolrd @siddarthpaim @j_bindra @adarshatva1 @mrityuandprem @alphonso269536 @pk00202500 @belliethomas @insignifictcman @generalusername @republic @thekrishnambiar @thedesignobsess @aravsrinivas @uidai @huggingface

Top Social Posts

Top posts by engagements in the last [--] hours

"Beyond just transcription Saaras V3 excels in three areas. It can automatically detect the language of the audio. It now comes with word-level time-stamps. And it also supports diarisation for multi-speaker audios. Try out your polyglot skills in the blog"
X Link 2026-02-11T15:16Z 14.4K followers, 12.4K engagements

"@aiexplorations Agree - there is much improvement to be done. Both in the schwa representation and in intonation. Happy to collaborate"
X Link 2026-02-09T04:49Z 13.4K followers, [--] engagements

"In a head-to-head comparison against leading AI dubbing platforms experts preferred Sarvam Studios outputs as the outright choice. The dubbed voice retains the tone pacing and expressiveness of the original speaker. For fine adjustments such as audio-visual alignment content teams can edit and regenerate specific segments directly in the Studio. https://twitter.com/i/web/status/2022251125781774450 https://twitter.com/i/web/status/2022251125781774450"
X Link 2026-02-13T10:06Z 14.4K followers, [----] engagements

"In a head-to-head comparison of real-world document translations Studio achieved the highest translation quality rating. Our outputs were recognised to have the most consistent quality across complex domains including academic fiction and legal content. Also Studio maintains the original structure and visual integrity such as preserving tables headings diagrams and page hierarchy without requiring manual redesign"
X Link 2026-02-13T10:06Z 14.4K followers, [----] engagements

"Drop 1/14: We pulled off a first of its kind. The budget speech by the Finance Minister was live telecast in multiple languages on @republic. Reached millions of homes with a latency under [--] minutes. https://www.youtube.com/watchv=wiQ9KesTBbA https://www.youtube.com/watchv=wiQ9KesTBbA"
X Link 2026-02-01T18:24Z 14.2K followers, 37.1K engagements

"Complex multi-column layouts of an old newspaper in Malayalam"
X Link 2026-02-05T13:52Z 14.1K followers, 11.4K engagements

"Odisha is positioning itself to become Indias compute backbone leveraging its unique position today of being both power and revenue surplus. Our partnership envisages building a 50MW AI-optimised facility that will function as the States AI public utility - powering e-governance agriculture advisories healthcare triage unified citizen helplines AI tutors and disaster management. This hub will also attract innovative companies into the state by providing access to cutting-edge AI infra thereby transitioning from being a mine-driven to a mind-driven economy."
X Link 2026-02-09T04:59Z 13.7K followers, 11.5K engagements

"First a little more elaboration on the results. We picked up a standard ETL task of extracting about [---] finance and business metrics from about [--] PDF documents pertaining to a companys performance across years. We want an agentic system to parse the documents and extract the metrics as per a given schema. Accuracy is measured as the percentage of the metrics correctly extracted. When we threw Claude Code at this problem we managed low accuracy values of about 17% at costs of about $12 for a run taking over [--] minutes. We then built a multi-agent system with Arya in six design iterations"
X Link 2026-02-10T06:57Z 13.8K followers, [----] engagements

"Now lets say a little more about our design motivation. Arya starts from a simple observation - current orchestration frameworks fail on two fronts. They lack runtime reliability with no infrastructure-grade guarantees on execution semantics inter-agent communication or state management. And they lack rapid iterability with no formal design language step-wise replay or modular observability. To address this we built Arya with four principles: Composable primitives: Every system is composed of only one of eight primitives: LLM Agent MCP Node Ledger Task Graph Code Interpreter and Artefact."
X Link 2026-02-10T06:57Z 13.7K followers, [----] engagements

"I go around saying we should be builders across the 'full stack' @SarvamAI - from compute to data to models to apps. And in come @thekrishnambiar and @thedesignobsess and say we should compose our own music for model launches And so here we go with our first Sarvam sound track. India's most ambitious decade to build is here; come join the build. And yes for your ringtone upgrade go here: https://www.sarvam.ai/bulbul-audio https://www.sarvam.ai/bulbul-audio"
X Link 2026-02-07T16:43Z 14.4K followers, 56.5K engagements

"Drop 7/14: Introducing Sarvam Arya - our multi-agent orchestration platform built from the ground up with robust systems engineering principles and frontier AI-assisted developer experience. We illustrate the Arya advantage on a common ETL task of extracting structured data from unstructured documents. Arya system with GPT [---] mini achieved 5x higher accuracy at 10x lower cost compared to Claude Code with agent swarm. Read more about Arya in this thread and our blog: https://www.sarvam.ai/blogs/introducing-sarvam-arya https://www.sarvam.ai/blogs/introducing-sarvam-arya"
X Link 2026-02-10T06:57Z 14.4K followers, 132.4K engagements

"On each of the most popular [--] languages on the IndicVoices benchmark Saaras V3 outperforms leading models such as Deepgram Nova-3 Elevenlabs Scribe v2 Gemini [--] Pro and GPT-4o Transcribe. The Realtime streaming model achieves much lower latency while being within a percentage point of the best error rate. https://twitter.com/i/web/status/2021604280583807054 https://twitter.com/i/web/status/2021604280583807054"
X Link 2026-02-11T15:16Z 14.4K followers, 14.8K engagements

"From lecture videos to national addresses from textbooks to novels Sarvam Studio is powering creation of multilingual content. We are excited to see what you will build. Read our blog for more details and to get started. https://www.sarvam.ai/blogs/sarvam-studio https://www.sarvam.ai/blogs/sarvam-studio"
X Link 2026-02-13T10:06Z 14.4K followers, [----] engagements

"Fully agree on the core necessity for affordable and efficient AI. We need to dramatically increase per capita access to intelligence without ballooning per capita energy costs. What Sarvam AI has demonstrated - world class AI can be done much more affordably and sustainably with much lower energy and compute footprint - is important. I have come to believe that in software code generation much more energy and compute efficient approaches to AI can be What Sarvam AI has demonstrated - world class AI can be done much more affordably and sustainably with much lower energy and compute footprint"
X Link 2026-02-14T02:47Z 14.4K followers, 15.1K engagements

"As much about our TTS model as is about the impeccable diction that Sanskrit affords. Code x Culture. Create these beautiful cards here - https://dashboard.sarvam.ai/text-to-speech Stress testing Bulbul v3 https://t.co/2hw4MUuBz9 https://dashboard.sarvam.ai/text-to-speech Stress testing Bulbul v3 https://t.co/2hw4MUuBz9"
X Link 2026-02-09T02:07Z 14.4K followers, 17.6K engagements

"Drop 9/14: Today we are introducing Sarvam Studio our product to help creators go multilingual. One piece of content every corner of India. With AI video dubbing Studio generates high-fidelity dubs in [--] Indian languages. In an expert study participants preferred Sarvam Studio for overall quality and production readiness. With agentic document translation Studio excels in contextually translating long-form content across genres. Our evaluations demonstrate that readers strongly preferred the output from Studio across different genres. https://twitter.com/i/web/status/2022251117359636843"
X Link 2026-02-13T10:06Z 14.4K followers, 227.1K engagements

"Drop 10/14: Announcing Sarvam Edge our dedicated effort to bring intelligence to run offline and on-device. Our goal is to make AI that is efficient private and accessible everywhere. Today we are showcasing our speech recognition speech synthesis translation and document digitisation models running on-device. This comes on the back of our research efforts to make models super small in memory and compute footprints while being close to accuracy of much larger models. We believe running models at the edge will truly unlock population scale AI in India. And we are working with partners to make"
X Link 2026-02-14T17:40Z 14.4K followers, 20.7K engagements

"Two weeks of daily drops of what we have been building at @SarvamAI starts today"
X Link 2026-02-01T02:30Z 14.4K followers, 68.9K engagements

"Drop 2/14: Sarvam Audio: a state-space based efficient audio language model that defines the new benchmarks in speech recognition for Indian languages. Significantly outperforms Gemini [--] and GPT 4o Transcribe in a range of benchmarks. See details in our blog: @SarvamAI https://www.sarvam.ai/blogs/sarvam-audio https://www.sarvam.ai/blogs/sarvam-audio"
X Link 2026-02-03T01:01Z 14.4K followers, 69.9K engagements

"Drop 4/14: Introducing Sarvam Vision: a state-space based [--] billion parameter vision language model that is competitive with the best results in digitisation in English and defines a significantly higher bar for Indian languages. See the details in our blog: https://www.sarvam.ai/blogs/Sarvam-vision https://www.sarvam.ai/blogs/Sarvam-vision"
X Link 2026-02-05T13:51Z 14.4K followers, 585.8K engagements

"Sarvam Vision achieves state-of-the-art accuracy of 84.3% on the olmOCR-Bench (English only subset) outperforming frontier models like Gemini [--] Pro and recent OCR models like DeepSeek OCR 2"
X Link 2026-02-05T13:51Z 14.4K followers, 94.9K engagements

"Older book with long form Tamil"
X Link 2026-02-05T13:52Z 14.4K followers, 11.4K engagements

"Drop 5/14: Introducing Bulbul V3 our latest text-to-speech model. It raises the bar for how human it sounds while being super robust. In an independent third-party human listening study Bulbul V3 delivers the highest listener preference and low error rates across use-cases and languages. See details in our blog but first watch the video. https://www.sarvam.ai/blogs/bulbul-v3 https://www.sarvam.ai/blogs/bulbul-v3"
X Link 2026-02-07T06:34Z 14.4K followers, 386.2K engagements

"Bulbul V3 is live. This is your month to go all in. Unlimited usage open through February. The mic is yours https://dashboard.sarvam.ai/text-to-speech https://dashboard.sarvam.ai/text-to-speech"
X Link 2026-02-07T06:34Z 14.4K followers, 11.2K engagements

"Drop 6/14: @SarvamAI is proud to announce a landmark in Indias sovereign AI journey through strategic partnerships with the Governments of Odisha and Tamil Nadu. The aim of these partnerships is to drive transformation by building at-scale compute sovereign models and the institutional capacity to drive AI adoption. Instead of going bottom-up with PoCs which find it hard to take off these partnerships will go top down with a state-wide mandate to deploy AI as a public utility embedding intelligence across all departments"
X Link 2026-02-09T04:59Z 14.4K followers, 54K engagements

"The advantages of setting up sovereign AI cannot be over-stated. Today we face a digital asymmetry. We are exporting data and importing tokens. When our data flows outward it trains foreign models and the learning loops compound elsewhere leaving us with tokens but no leverage. States that move quickly will build enduring advantage. A case in point is the U.S. wherein AI drove nearly 3% GDP growth in [----]. If states in India can generate even 1% annual growth via AI the compounding effect towards [----] can be massive"
X Link 2026-02-09T04:59Z 14.4K followers, 27.5K engagements

"The other key insight that Arya is based on is the following. Today there are two ends of the spectrum in agentic workflows. Simple LLM chains that run autonomously but handle only straightforward tasks. And frontier coding agents that tackle complex problems but require heavy human interaction. The missing middle is production-grade agents - systems that are genuinely sophisticated but run at scale with reliability thousands of times a day without a human in the loop. For this missing middle of agents frontier models can be moved from runtime to compile time. Like compilers they can"
X Link 2026-02-10T06:57Z 14.4K followers, [----] engagements

"We plan to open-source Arya as this layer should be shared infrastructure the way databases and container orchestration are. This release will include a fully containerised runtime that can run anywhere a debug/replay user-interface for reviewing runs and an LLM-powered developer experience for authoring and deployments. Code coming soon. Meanwhile we are building for our enterprise partners with Arya. https://www.sarvam.ai/blogs/introducing-sarvam-arya https://www.sarvam.ai/blogs/introducing-sarvam-arya"
X Link 2026-02-10T06:57Z 14.4K followers, 14.9K engagements

"India's decade to build with the values we stand for The all-new Sarvam is here. Designed for all of India. https://t.co/O78H03Jcki https://t.co/zxtycbcDAu The all-new Sarvam is here. Designed for all of India. https://t.co/O78H03Jcki https://t.co/zxtycbcDAu"
X Link 2026-02-11T04:01Z 14.4K followers, 54.3K engagements

"Drop 8/14: Introducing Saaras V3 the next iteration of our speech recognition model. We extend our lead in this space with an even more accurate model particularly for mixed-language and noisy speech. We have also expanded support for all the [--] scheduled languages of India. And we now support real-time streaming In this mode the model delivers low latency while preserving transcription quality. https://www.sarvam.ai/blogs/asr https://www.sarvam.ai/blogs/asr"
X Link 2026-02-11T15:16Z 14.4K followers, 145.8K engagements

"The streaming mode is fun: you can literally use it with a movie playing in the background Try it out here: https://www.sarvam.ai/blogs/asr#asr-sarvam https://www.sarvam.ai/blogs/asr#asr-sarvam"
X Link 2026-02-11T15:16Z 14.4K followers, [----] engagements

"Whether you are building voice bots subtitling long audios or analysing call transcripts you would find value in building with Saaras V3. Get started with our APIs today: https://dashboard.sarvam.ai/speech-to-text https://dashboard.sarvam.ai/speech-to-text"
X Link 2026-02-11T15:16Z 14.4K followers, [----] engagements

"This video shows document intelligence on device. An Odia document is digitised entirely on a MacBook Pro. We believe a range of such models will come bundled in productivity focused laptops"
X Link 2026-02-14T17:40Z 14.4K followers, [----] engagements

"In this second example we demonstrate a hybrid edge-cloud architecture applied to a stock broking app. The user can fetch live data query a portfolio execute trades and review research all with voice processing running locally and LLMs and APIs on the cloud. We believe apps will be able to provide convenient hands-free voice experiences at super low operating costs. https://twitter.com/i/web/status/2022727772632998131 https://twitter.com/i/web/status/2022727772632998131"
X Link 2026-02-14T17:40Z 14.4K followers, [----] engagements

"In this demo we show real-time voice translation between Indian languages. The user can speak in their preferred language and the system translates and responds in another preserving both meaning and natural delivery. Speech recognition translation and expressive text-to-speech work together seamlessly"
X Link 2026-02-14T17:52Z 14.4K followers, [----] engagements

"Sarvam Edge is being developed in close collaboration with leading global device manufacturers combining Sarvam's models with years of hardware R&D to power the next generation of consumer devices. If you are an app developer who would like to use our Edge SDK or a device manufacturer who would like to bundle our models then reach out at edge@sarvam.ai https://www.sarvam.ai/blogs/sarvam-edge https://www.sarvam.ai/blogs/sarvam-edge"
X Link 2026-02-14T17:53Z 14.4K followers, [----] engagements

"@AravSrinivas Aravind at @SarvamAI we are building sovereign models that combine deep reasoning and Indic language skills. Would love to have you join this mission"
X Link 2025-01-22T13:13Z [----] followers, 114.7K engagements

"Next week I am in the Bay Area looking to meet exceptional researchers in the AI space. Come see what we have been building at @SarvamAI and what you can do to innovate for a billion Indians RSVP with a DM"
X Link 2025-03-12T17:59Z [----] followers, 22.2K engagements

"Delighted to be working with @UIDAI in bringing AI to one of the most crucial tech infrastructures of our country"
X Link 2025-03-19T06:10Z [----] followers, 13.7K engagements

"India needs to be a builder and not just a consumer in this defining era of technology. The journey towards strategic autonomy starts now DMs open. It's official We're proud to announce that Sarvam has been selected by the Government of India under the IndiaAI Mission to build India's sovereign Large Language Model. Building India's sovereign model from the ground up is a crucial step toward Atmanirbhar Bharat. The https://t.co/nA4WgQ0aw0 It's official We're proud to announce that Sarvam has been selected by the Government of India under the IndiaAI Mission to build India's sovereign Large"
X Link 2025-04-26T10:49Z [----] followers, 93.6K engagements

"Clarification: This is not a grant. A gov body will take equity in Sarvam for the compute we receive. And we are committed to building public interest use-cases and enabling the ecosystem in various ways such as hyper-optimising the inferencing costs in India. BTW we as a country have a long way to go in AI. This is the time to strengthen each other. Let's go"
X Link 2025-04-27T05:48Z [----] followers, 63K engagements

"Great to be receiving feedback on Sarvam-M. Please keep them coming. Will help strengthen our pipelines as we start to train our sovereign model. This was particularly interesting - https://alokbishoyi.com/blogposts/sarvam-political-evaluation.html https://alokbishoyi.com/blogposts/sarvam-political-evaluation.html"
X Link 2025-05-25T07:46Z [----] followers, 58.4K engagements

"Good to see folks building creatively with Sarvam-M which is now trending on @huggingface 🤗 More coming soon"
X Link 2025-05-28T08:28Z [----] followers, 38.9K engagements

"Was a fun chat with @mukeshbansal06 on LLMs strategic autonomy what Indian businesses can do in AI and consciousness vs intelligence - https://youtu.be/89ZqoXfdGT4 https://youtu.be/89ZqoXfdGT4"
X Link 2025-06-11T15:43Z [----] followers, 14.8K engagements

Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing

creator/x::pratykumar
/creator/x::pratykumar