Models
Sarvam AI provides a purpose-built AI stack for building applications in Indian languages. Our models span speech-to-text, speech translation, text translation, and high-quality text-to-speech—designed specifically for India’s linguistic diversity, accents, and real-world usage patterns.
Each model is trained and evaluated on Indian languages and culturally grounded data, enabling higher accuracy in production scenarios. With simple, well-documented APIs and predictable performance, developers can build, deploy, and scale India-first AI experiences without managing model complexity.
New to building for Indian languages? Start with Building for Indian Languages — a practical guide to language coverage, code-mixing, scripts, native numerals, 8kHz telephony audio, and pronunciation control.
Model Selection Guide
Sarvam models: Saaras v3 — Speech to Text, Bulbul v3 — Text to Speech, Mayura — Text Translation, Sarvam-Translate — Extended Translation, Sarvam-105B — Flagship Chat LLM, Sarvam Vision — Document Intelligence.
Sarvam also serves a set of open-source models, listed separately below.
State-of-the-art ASR with 23 language support (22 Indian + English) and multiple output modes: transcribe, translate, verbatim, translit, codemix.
Natural-sounding voices for 11 languages (10 Indian + English) with customizable pitch, pace, and speaker options.
High-quality translation between 11 languages (10 Indian + English) with context preservation.
Extended translation support for all 23 languages (22 Indian + English) with superior accuracy.
105B parameter flagship model — sarvam-105b for complex reasoning and agentic tasks, plus sarvam-105b-conversations for real-time dialogue and voice agents.
Extract and digitize content from documents in 23 languages with accurate OCR and structured output.
Open-Source Models
Sarvam also serves a small set of open-source models, reachable with the same API key and credits. They are not tuned for Indian languages — use these only when you need a capability Sarvam’s own models do not offer, and prefer a Sarvam model for anything involving Indian languages, Indic scripts, or code-mixed input.
GLM-5.2 and Gemma 4 31B are available in beta and rolling out gradually. See Access to Beta APIs.
Open-source models are on the OpenAI-compatible /v2/chat/completions endpoint,
alongside sarvam-105b. Sarvam chat models — including sarvam-105b-conversations —
are on /v1. See
using an open-source model.
512K token context window with tool calling and visible reasoning — for tasks that exceed the 128K context of Sarvam’s chat models. Text only.
Image input and tool calling, for visual and classification tasks.
Read more about the open-source models →
Language Support Overview
Language coverage varies by model — check the table below before picking one. Full per-model tables are linked from each model’s own page.
23-language set (Saaras v3, Sarvam Translate, Sarvam Vision)
11-language set (Bulbul v3, Mayura, Sarvam-105B, Saarika v2.5)
Use Cases
Voice Assistant
Content Localization
Call Center Analytics
Educational Platform
Document Processing
Build a multilingual voice assistant
- Speech Input: Use Saaras v3 with
mode="transcribe"to convert user speech to text - Understanding: Process with Sarvam-105B for intelligent responses
- Speech Output: Convert responses to speech with Bulbul
Perfect for customer service, smart home devices, and accessibility applications.
Legacy Models
The following models are still available but are being phased out. We recommend migrating to the newer models listed above.
24B parameter multilingual chat model with hybrid reasoning. Deprecated and no longer available through the API — migrate to Sarvam-105B.
30B parameter multilingual chat model with strong reasoning and conversational capabilities. Deprecated — migrate to Sarvam-105B.
Legacy ASR model supporting 11 languages (10 Indian + English). Migrate to Saaras v3 for improved accuracy and 23-language support.