I am using Futo Keyboard at the moment (Gboard is not an option I would like to use). If I am not missing something, there have been no updates / improvements in voice recognition since its release: the pages on voice recognition models on its website are still labeled "coming soon / under construction". As far as I understand, transcription is still based on whisper-small, released in 2022)
AI has done giant leaps in the past few years; is this model still the state of the art? Have whisper updates, or better open-source (or at least offline source-available / open weights) voice recognition engines appeared in the past months/years?