$0.002 per minute.
There is no free API; it costs $0.002 per minute. The weights are open, so you can run it yourself.
#33 of 51 in Speech to text.
On AA Speech to text · Recorded audio word error rate, Parakeet TDT 0.6B V3, Togetherai scores 4.5 vs 4.5 — about the same. It is 50% cheaper.
Released 2025-08-04. It is the current version in its line.
Yes, the weights are open (cc-by-4.0 licence).
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Parakeet TDT 0.6B V3, Togetherai is an open-weights speech & voice model from NVIDIA, released Aug 2025.
Modulate STT Batch English VFast (cheaper), MAI Transcribe 2 (better), Whisper Large v2 (open weights).