From day one, Jivam supports all 22 languages scheduled under the Indian constitution. This isn’t a feature we added — it’s a property of the model we chose.
Sarvam-105B was trained on native Indian language data at a scale that no other publicly available model has matched. The result is genuine multilingual fluency — not translation, not approximate understanding, but the ability to reason, draft, and respond in the language you’re thinking in.
The 22 languages
Assamese, Bengali, Bodo, Dogri, Gujarati, Hindi, Kannada, Kashmiri, Konkani, Maithili, Malayalam, Manipuri, Marathi, Nepali, Odia, Punjabi, Sanskrit, Santali, Sindhi, Tamil, Telugu, and Urdu.
Every one of these is supported natively in Jivam. You can write in any of them, switch between them mid-conversation, and Jivam will follow along.
Why it matters
Most AI tools handle English well and everything else poorly. They’ll respond to a question in Hindi, but the quality drops noticeably. The reasoning feels shallower. The response feels translated rather than native.
With Sarvam-105B, this isn’t the case. The model was trained to think in Indian languages, not to translate into them. When you ask a question in Tamil, the model reasons in Tamil — it doesn’t convert your question to English, think in English, and translate back.
How to use it
Just start writing in the language you prefer. Jivam will respond in kind. You can switch languages at any point in the conversation and Jivam will follow. There’s no language setting to configure — it just works.