isitsota / categories / Speech-to-text
Speech · Oct 14, 2026
Who is SOTA in Speech-to-text?
Nobody yet.
English short-form word error rate on the Hugging Face Open ASR Leaderboard. Lower is better. A tight top of the board does not get a forced crown. As of Oct 14, 2026, Speech-to-text has no model with enough independent evidence for a SOTA crown.
No open model on the current board has a known parameter count at or below 32B. Counts come from Epoch, or from a size written in the model name, such as 27B.
Standings · le-32b
The board
| # | Model | State | Confidence |
|---|
Evidence recipe
What can move this crown
- Open ASR Leaderboard (English short-form) — primary, weight 1