Sarvam AI released Saaras V4 speech recognition model supporting 22 Indian languages

Sarvam AI released Saaras V4, a speech recognition model that supports 22 Indian languages and English and can output five transcript modes (verbatim, normalised, code-mixed, transliterated, translated) from the same audio input. The company said the model supports real-time streaming ("time to first token below 150 ms") and is available via Sarvam AI's API with Python and Node.js SDKs and integrations including Vercel AI SDK, LiveKit Agents and Pipecat Agents.