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
- Micius-Flash: 27B Multimodal Model for Hardware and Semiconductor Agents — zhiman-ai · 2026-11-11
- Baidu says RSI auto-optimized ERNIE inference, 6.24x decode speedup on Kunxin P800 — xiaohu · 2026-11-11
- Baidu's ECHO: multi-token prediction reusing the last layer as a drafter — xiaohu · 2026-11-11
- ERNIE 5.5 Flash Preview: 280B-param multi-router MoE activating only 14B per token — xiaohu · 2026-11-11
- ERNIE 5.5 recreates the Pelican test: 117-second animation drawn frame-by-frame in Python — xiaohu · 2026-11-11
- 35,000+ Grok-built apps exposed patient data, plaintext passwords, and sales pipeline — The Decoder · 2026-11-11