Sarvam launches Saaras V4 ASR model covering 22 Indic languages with sub-150ms streaming
itsOmSarraf_ · x · 2026-09-25
Indian AI startup Sarvam released Saaras V4, an ASR model built for noisy, accented, dialect-heavy and mixed-language speech.
- Covers 22 Indic languages plus English, claiming SOTA across all of them
- Five modes: Transcribe, Translate, Transliterate, Codemix, Verbatim
- Keyterm prompting steers the model toward domain-critical words
- Real-time streaming with <150ms time-to-first-token
- Available via API and inside Sarvam's voice agent orchestration stack
More from Models
- Embedding test: RAG ranks the no-refund policy first, showing models still matter — galratner · 2026-09-25
- Meme: doing your taxes with Mistral could cost you a $20k IRS visit — Rasmic · 2026-09-25
- Users rave about Anthropic Opus 5.5: Fable-level smarts for 40% less, hard to burn the quota — altryne · 2026-09-25
- Industry voice: RL environments are largely generated at scale by closed models, not human labs — xeophon · 2026-09-25
- Opus 5.5 Review: Every.to Says It's Pulling Codex Converts Back to Claude — every · 2026-09-25
- Creator endorses model routers as the sensible answer to model switching fatigue — Rasmic · 2026-09-25