While I'm sure the new Gemini model has made improvements, I feel like the user experience outside of the model itself is stagnating. I think OpenAI's interfaces, both web app and mobile app, are quite a bit more polished currently. For example, Gemini's speech recognition struggles with longer pauses and often enough cuts me off mid-sentence. Also, OpenAIs whisper model understands more context (for instance, saying “[...] plex, emby and Jellyfin [...]” is usually understood in whisper, but less often in Gemini) The Gemini web app lacks keyboard shortcuts for basic actions like opening a new chat or toggling the sidebar (good for privacy friendly pair programming). Last point off the top of my head would be the ability to edit messages beyond just the last one. That's possible in ChatGPT, but not in Gemini. Googlers are spending so much money for model training, I would appreciate spending some for making it fun to use :)