Hi everyone!
I hope you’re doing well.
- Modalities: Gemma 3 supports text, image, and video inputs. Audio is not supported. For AI Studio, the team is working in adding them. Locally, you can use all of them.
- System instruction: the model does not support a system role, but passing the system prompt as the first turn works very well. See Gemma formatting and system instructions | Google AI for Developers
- Code execution: we’re not planning to add code execution for now, but we’ll discuss it!
- Function calling: this also relies on the instructability of the model. See Function calling with Gemma | Google AI for Developers
Thanks everyone for the feedback, we’ll iterate in our documentation for this.