Previously:

Sept 2025 https://news.ycombinator.com/item?id=45114245

That was about the text-to-speech model, the speech-to-text one was release in January.