{"version":1,"type":"story","url":"https://digestai.news/story/sarvam-ai-launches-saaras-v4-speech-to-text-model-for-22-indian-langua","json":"https://digestai.news/story/sarvam-ai-launches-saaras-v4-speech-to-text-model-for-22-indian-langua.json","markdown":"https://digestai.news/story/sarvam-ai-launches-saaras-v4-speech-to-text-model-for-22-indian-langua.md","slug":"sarvam-ai-launches-saaras-v4-speech-to-text-model-for-22-indian-langua","headline":"Sarvam AI launches Saaras V4 speech-to-text model for 22 Indian languages and global English","summary":"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 showing lower word error rates (WER) than competitors like Deepgram Nova-3 and GPT-4o Transcribe in some cases.\n\nKey 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.","keyPoints":["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"],"whyItMatters":"This model fills a critical gap for multilingual speech recognition in India, where English and regional languages coexist. Its low-latency streaming and multi-mode outputs could simplify workflows for call centers, media, and accessibility tools, though vendor claims await independent verification.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["Sarvam AI","MarkTechPost","Deepgram","OpenAI","AI4Bharat","Hugging Face"],"models":["Saaras V4","Saaras v3","GPT-4o Transcribe","Deepgram Nova-3","Svarah","IndicVoices"],"people":["Asif Razzaq"]},"firstPublishedAt":"2026-09-26T21:56:23Z","updatedAt":"2026-09-26T21:56:23Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"MarkTechPost","title":"Sarvam AI Releases Saaras V4: A Speech-to-Text Model for All 22 Indian Languages and Global English","url":"https://marktechpost.com/2026/09/26/sarvam-ai-releases-saaras-v4-a-speech-to-text-model-for-all-22-indian-languages-and-global-english","publishedAt":"2026-09-26T21:56:23Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":{"title":"Rapid Advances in Multilingual Speech Recognition Technology","url":"https://digestai.news/thread/xai-releases-grok-voice-transcribe-2-0-speech-to-text-model","storyCount":2},"cite":{"text":"Digest AI, \"Sarvam AI launches Saaras V4 speech-to-text model for 22 Indian languages and global English\", 26 September 2026, https://digestai.news/story/sarvam-ai-launches-saaras-v4-speech-to-text-model-for-22-indian-langua","publisher":"Digest AI","title":"Sarvam AI launches Saaras V4 speech-to-text model for 22 Indian languages and global English","datePublished":"2026-09-26T21:56:23Z","url":"https://digestai.news/story/sarvam-ai-launches-saaras-v4-speech-to-text-model-for-22-indian-langua"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}