Table of Contents
Bengaluru Startup Launches Two New AI Models
Introduction
- Company Announcement – Sarvam AI introduced two new AI language models called Sarvam-30B and Sarvam-105B.
- Launch Event – The announcement was made during the India AI Impact Summit 2026 where government officials, technology experts and business leaders were present.
- Launch Date – The models were officially launched in February 2026.
- Indian Vision – The launch supports India’s idea of “sovereign AI,” which means building AI systems inside the country for Indian needs.
- National Importance – This step helps India depend less on foreign technology companies and build its own strong AI system.
About Sarvam AI
- Bengaluru Location – The company is based in Bengaluru and works mainly in artificial intelligence research.
- Indian Focus – It builds AI tools that understand Indian languages and local requirements.
- Big Goal – The company wants to compete with large international AI companies.
- Growing Sector – Its work supports the fast growth of India’s AI industry and skilled workforce.
Sarvam-30B Model Overview
- Model Size – Sarvam-30B has 30 billion parameters, which are like small parts that help the model learn and think.
- Smart Structure – It uses a mixture-of-experts system, meaning it does not use all parts at once but only the necessary ones.
- Less Power Use – Around 1 billion parameters are active at one time, which saves computing cost and electricity.
- Long Memory – It can handle up to 32,000 tokens at a time, so it can understand long conversations and documents.
- Large Training Data – It was trained on nearly 16 trillion words and data pieces to improve its understanding and answers.
- Daily Use – It is suitable for chatbots, question-answer systems, and coding help.
- Language Support – It works well in many Indian languages, not just English.
Sarvam-105B Model Overview
- Bigger Model – Sarvam-105B has 105 billion parameters, making it more powerful for complex tasks.
- Very Long Context – It can handle up to 128,000 tokens at once, allowing it to study very long reports and documents.
- Deep Thinking – It is designed for advanced reasoning, research work, and business analysis.
- Public Demo – During the summit, it was shown analysing company financial data in real time.
- Strong Performance – The company says it performs as well as, or sometimes better than, some well-known global AI models.
- Business Use – It is mainly meant for companies, institutions and government work.
Technical and Strategic Features
Efficient System
- Expert Method – Both models use the mixture-of-experts method to improve speed and reduce waste of resources.
- Lower Cost – By using only needed parts at a time, they reduce running costs.
- Balanced Design – They offer strong performance without requiring extremely high computing power.
Multiple Languages
- 22 Languages – The models aim to support all 22 official Indian languages.
- Wider Access – This allows people who do not speak English fluently to use AI tools easily.
- Phone Friendly – The system was shown working on smartphones as well as basic feature phones.
- AI Assistant – A multilingual chatbot named “Vikram” was demonstrated during the event.
Position in Global AI Race
- Global Challenge – The launch shows India’s intention to compete with companies like OpenAI, Google and Anthropic.
- Model Comparison – The 105B model is said to perform competitively with systems such as Gemini Flash and DeepSeek R1.
- Affordable Option – It is expected to offer similar performance at a lower operating cost.
- Self-Reliance – This move strengthens India’s independence in advanced technology.
Significance for India’s AI Ecosystem
- Government Support – The launch matches the goals of the IndiaAI Mission, which promotes Indian-made AI systems.
- Data Control – It allows AI systems to be built and managed under Indian laws and rules.
- Industry Benefit – Sectors like finance, healthcare, education and research can use these models.
- Developer Support – Indian developers can build new apps and services using these homegrown AI tools.
Important Questions
- What are the main features of Sarvam-30B and Sarvam-105B models launched by Sarvam AI?
- How does the mixture-of-experts system in Sarvam-30B and Sarvam-105B reduce computing cost and improve efficiency?
- In what ways do Sarvam-30B and Sarvam-105B support multiple Indian languages and improve accessibility across India?
- How do Sarvam-30B and Sarvam-105B compare with global AI systems developed by OpenAI and Google?
- What is the importance of Sarvam-30B and Sarvam-105B for India’s sovereign AI vision and the IndiaAI Mission?
Conclusion
The launch of Sarvam-30B and Sarvam-105B by Sarvam AI represents a major development in India’s effort to build sovereign artificial intelligence under the framework of the IndiaAI Mission. The advanced features, mixture-of-experts system, multilingual support, and competitive performance of Sarvam-30B and Sarvam-105B strengthen India’s technological self-reliance and enhance India’s position in the global artificial intelligence landscape.
Download Online Mock Test Mobile APP
3
