Please restore gemini-2.0-flash-exp to AI Studio! It looks like it is still in Vertex but that doesn’t work with Live and gemini-2.0-flash-exp was pulled last night from AI Studio and now is no longer available. What is going on Google? gemini-2.0-flash-exp is the only model that supports text modality and Live Versions are not working in Vertex, this is so frustrating!
1 Like
@Seth_Ford understand the frustration. Just to clarify
- when you say AIS, are you referring to being able to use them from the Gemini Developer API?
- would you tell me a little bit about your use case
The whole Live API is a dumpster fire and where it can and can not be used and what model of live supports what be it AI studio or Vertex. Please read the following thread all the way to the bottom gemini-live-2.5-flash-preview doesn't call tools · Issue #1698 · google/adk-python · GitHub The EXP model is the only model that supports the Live API with Text modality on AI Studio and it was removed last night without notification, after being told it was going to stick around. Today I had to figure out if there is an actual Live model somewhere else that support STT aka text modality rather the STS on Vertex. It does look like gemini-2.0-flash-live-preview-04-09 does but then any rev past that does not? I do not want a full cascade aka STS as the only option in Live unless… You Support something Like this. Note… This documentation is incorrect Gemini Live API plugin | LiveKit Documentation Text Modality does not work with gemini-2.5-flash-native-audio-preview-12-2025 only audio does and you are locked into the few voices Google provides and there is no Ability to do any deep customization with the voice. Like Real-time TTS API with AI laughter and emotion | Cartesia Sonic-3 allows us to do.
1 Like
I was literally just about to post about this, I was using live api for the gemini 3 hackathon and I had to dig deep to find this model as it’s the only one that allowed text modality in live api despite flash audio being advertised as such. Was really disappointed when I woke up today and it didn’t run
1 Like
Yeah… And don’t even get me started on tool calling with these Live Modals or being able to provide hints / guidance back to the LLM during the session. The work arounds I’ve had to do to get these models to perform and sound human and call tools is crazy and has been months of time wasted. I’m so close to something amazing…
I’m wondering does gemini-2.0-flash-live-preview-04-09 work on your end for text modality?
Yes it does, it’s the only one that does though…
If you don’t mind me asking are you using vertex or gem sdk? Would you mind sharing your code? I tried to swap out exp for that model and I got the same error as I did with exp (I’m not on vertex).
I had to switch to Vertex as it is the only place it run. I have no open sourced my code at this time as it’s a mono repo.
1 Like