Sarvam-105B Is Here: India’s Biggest AI Leap Yet?

February 19, 2026 Harshvardhan Mishra

India has officially entered a new phase in the global artificial intelligence race.
On 18 February 2026, Indian startup Sarvam AI launched two powerful open-source large language models:

Sarvam-30B
Sarvam-105B

The announcement is being seen as a major step toward India building its own AI ecosystem instead of depending heavily on foreign technologies.

Why This Launch Matters

For years, most AI tools used in India were built outside the country. That created concerns around:

Data privacy
High API costs
Limited Indian language support
Dependence on foreign infrastructure

Now India is pushing for sovereign AI — AI systems trained, hosted, and controlled within the country.

The launch aligns with the government’s
👉 IndiaAI Mission
which aims to develop domestic foundation models and digital infrastructure.

Sarvam-30B: Efficient Enterprise AI

Sarvam-30B is designed to be smaller, faster, and cheaper while still powerful.

Key Specifications

Pre-trained on 16 trillion tokens
32,000 token context window
Optimized for lower compute cost

Use Cases

Customer support chatbots
Document summarization
Workflow automation
Enterprise knowledge assistants

This model targets companies that want AI but cannot afford extremely large compute resources.

Sarvam-105B: The Flagship Model

Sarvam-105B is the main attraction — a massive large language model built for complex reasoning and long conversations.

Key Capabilities

128,000 token context window
Handles long documents and detailed analysis
Comparable to global large-scale models
Built for enterprise-grade AI systems

Practical Applications

Legal & policy document analysis
Research assistants
Coding copilots
Government knowledge systems
Advanced multilingual chatbots

Fully Open-Source Approach

A major highlight:
Both models are open source.

This means:

Developers can modify them
Startups can build products without huge API fees
Governments can host them locally
Sensitive data stays inside India

Sarvam AI also emphasized that the models were trained from the ground up, instead of relying heavily on external datasets — important for data sovereignty.

Real-World Demo: AI on a Feature Phone

Sarvam showcased a chatbot named “Vikram” running on a basic keypad phone.

The assistant could converse in:

Hindi
Punjabi
Other Indian languages

This demonstrated a focus not just on premium smartphone users, but India’s mass population, including rural and low-internet users.

Why Local Language AI Matters

India has:

22 official languages
Hundreds of dialects
Millions of non-English users

Most global AI models are English-centric.

Sarvam-105B aims to solve:

Government service accessibility
Rural digital adoption
Education inclusion
Voice-based computing

Impact on India’s AI Ecosystem

For Startups

Lower cost → faster innovation → local AI apps

For Government

Secure hosting → sensitive data protection

For Enterprises

Custom private AI models without foreign dependency

For Citizens

AI tools in native languages

Global Implications

India joining the foundation-model race means the world AI ecosystem may no longer be dominated only by US and China.

Instead we may see:

Regional AI ecosystems
Language-specific models
Localized computing infrastructure

Conclusion

With Sarvam-30B and Sarvam-105B,
Sarvam AI has positioned itself as a serious contender in the global AI space.

By making the models open source, supporting Indian languages, and aligning with national infrastructure goals, India has taken a major step toward technological independence.

Sarvam-105B may not just be another AI model — it could mark the beginning of India’s sovereign AI era.