Sarvam AI launches Saaras V4 speech-to-text model for 22 Indian languages and global English
Sarvam AI released Saaras V4, a speech-to-text model supporting all 22 scheduled Indian languages plus global English accents. The encoder-decoder system uses a 3B-parameter hybrid state-space language model (Sarvam-3B) trained in-house to process audio embeddings into transcripts. It claims state-of-the-art accuracy across languages, including noisy audio, with vendor-reported benchmarks…
Key points
- Saaras V4 supports 22 Indian languages and global English accents in one model, with vendor-reported state-of-the-art accuracy
- Keyterm prompting allows up to 50 terms to bias transcription, and five output modes let users choose transcription style
- API pricing is ₹30/hour for real-time transcription, with streaming support and sub-150ms latency
Key features include five output modes (transcription, verbatim, codemix, transliteration, and translation), keyterm prompting (up to 50 terms), and streaming capabilities with sub-150ms time-to-first-token. Pricing starts at ₹30 per hour for real-time transcription, with optional diarization at ₹45/hour. The model is available via Sarvam’s API today, though self-hosting documentation still references Saaras v3. No independent validation of the benchmarks has been published yet.
Sarvam AI Releases Saaras V4: A Speech-to-Text Model for All 22 Indian Languages and Global English
MarkTechPost · 26 September 2026
Loading the full article…
This text was published by MarkTechPost and written by Asif Razzaq. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Apiiro finds 1,122 public GitHub files with secrets saved by Claude Code · 1 src
- Thinking Machines releases Inkling 2 Small, a 594B MoE model with 17B active parameters · 8 src
- OpenAI releases GPT-6 Cyber, a Critical-rated model that outperforms predecessors in vulnerability detection and alignment · 25 src
- Google releases Gemini Omni 1.2 Flash with 15-second video generation and 30% price cut to $0.07 per second · 10 src
- US accuses Chinese executive of allegedly sharing Nvidia chip info with Beijing · 11 src
Comments
via GitHub Discussions