Skip to content

Speech recognition

Open ASR Leaderboard / ESB

Speech transcription across multiple domains, with multilingual and long-form tracks.

About Open ASR Leaderboard / ESB

What it measures

Normalized WER plus speed/throughput under a declared protocol.

Scope & limitations

The suite is not a single uniform corpus license. Normalize transcripts identically and separate long-form, multilingual and English tracks.

Evaluation environment

Speech-to-text API or local model.

Source

Original benchmark

Dataset & task catalog ↗