New Models needed

Hey, some feedback on the model side of things. AI moves fast, and the models Blabby is using for the second "mode" layer are starting to feel dated. There are newer, cheaper, faster, smarter options out now. To be clear, I still like Blabby better, but even Codex just added transcription to their desktop app running the newer stuff, with global hotkey activation, so the gap is closing.

For the AI modes/post-processing layer, gpt-5.6-luna as the fast default and gpt-5.6-terra as the higher-quality option, plus Gemini 3.6 Flash and 3.5 Flash-Lite. Honestly, I'd love to see Gemini 3.6 Flash as it just came out and it's extremely token-efficient.

For the initial transcription layer, nothing really beats Whisper V3 Large on Groq for speed and cost (~$0.11/hr at waaay past real-time). However, for a more premium option/tier, ElevenLabs Scribe v2 makes roughly half the errors Whisper does. It is 3x the price, but I think plenty of users would happily pay a bit extra for the increased accuracy.

Would love to see Blabby stay ahead here. Thanks for the great service!