Audio benchmarks
Benchmark-specific speech recognition results. The current dataset does not define a cross-benchmark overall speech score.
Data as of 2026-08-07
Each chart preserves its original metric and ranking direction.
No single cross-benchmark winner is calculated. Use each benchmark leader above because the evaluations can use different datasets and metrics.
The current registry preserves each speech benchmark's original metric and direction instead of combining unlike results.
This page currently contains 4 published speech-to-text benchmark definitions.