Sarvam Prism
Decision model built on Sarvam 105B. Picks from a fixed set of options and returns calibrated probabilities, in about 44 ms.
Sarvam 105B Conversations
Sarvam 105B post-trained for real-time dialogue, voice agents, and multi-turn conversations in 10 Indian languages and English.
Bulbul V4 Flash
Expressive, native text-to-speech across 11 languages, with 83 voices in 224 styles and first audio in under 250 ms.
Sarvam Vision 2.1
Document intelligence model that parses complex tables, extracts key-value pairs from forms, and reads handwritten text in Indian languages.
Saaras V4
Speech recognition for 22 Indian languages and English, with five output modes including transcribe, translate, and codemix.
Sarvam 105B
Flagship large language model delivering state-of-the-art performance across Indian languages for enterprise-grade applications.
Saaras V3
Streaming speech recognition for 22 Indian languages with low-latency decoding and code-mixed support.
Bulbul V3
Natural, expressive, and production-ready text-to-speech for Indian languages across 11 languages.
Sarvam Vision
3B state-space vision-language model for document parsing, OCR, and visual understanding in Indian languages.
Sarvam Translate
Open-weights translation model supporting 22 Indian languages with structured long-form text.
Mayura
Translates directly between English and 10 Indian languages, and reads code-mixed input as-is.
Sarvam Voice Agents
Turn Sarvam 105B into a voice agent
Calls in 11 Indian languages, as low as ₹3 a minute.
Start building with our models Explore, test, and deploy
Start building with our models
Explore, test, and deploy