Sarvam Indic DiarBench

Overview

Sarvam AI, in collaboration with AI4Bharat, introduced Indic DiarBench as a benchmark dataset and evaluation suite for joint automatic speech recognition (ASR) and speaker diarization across 22 scheduled Indian languages.[4] It standardizes testing for multi-speaker, multilingual Indian audio, enabling researchers and enterprises to compare ASR and diarization systems under realistic conditions.[4] The benchmark strengthens the tooling ecosystem around Indic speech technologies and supports the deployment of Sarvam’s own speech models in production settings.