Fact · Languages and accents
120million parameters
AI4Bharat's Hindi IndicConformer model is based on a Conformer-Large architecture with 120 million parameters and a hybrid CTC-RNNT decoder.
Checked against its source on October 17, 2026
In the source's own words
“It is based on a Conformer-Large architecture with 120M parameters and a hybrid CTC-RNNT decoder”
AI4Bharat- Hindi - IndicConformer Automatic Speech Recognition (ASR) Model (AIKosh model listing), AI4Bharat, AIKosh, the Government of India AI models portal, n.d., Model description.
Open the original- What it measures
- The model takes 16 kHz mono audio and outputs text. It is a model for developers who run speech recognition themselves, not a dictation app.
- What it does not show
- The listing gives no word error rate. You need to test it on your own Hindi audio before you rely on it.