On both nodes corresponding to Gemini model calls as I have tried, those being the Vertex AI node and the Gemini Model node, there is no option to support the new thinking_level parameter of Gemini 3 Flash.
Albeit since this model just released this is not frustrating, just food for thought so I can experiment with this model and its reasoning speeds.
thinking_level will not be of type number anymore, as an example of thinking_level will be “minimal”
see these docs for full details : Thinking | Generative AI on Vertex AI | Google Cloud Documentation Gemini 3 Flash | Generative AI on Vertex AI | Google Cloud Documentation
The idea is:
Support for Gemini 3 model thinking parameters
My use case:
RAG
I think it would be beneficial to add this because:
Is there an update or a rough ETA on when the UI will support the minimal, low, medium, and high dropdowns? It would be a huge help for those of us moving our production agents to Gemini 3 Flash/Pro.
+1 for this, really looking forward to using Gemini 3 models with Vertex once they are GA so this will be very useful when the time comes in the near future.
Since it is very slow on their side to validate/follow update of every AI nodes, the best is to create a community node your own or find one that does the job. I’m actively adding/using this node in production, any help would be great.
For my use cases, I added:
Labels (costs control)
Thinking Level
For the nodes: Vertex AI Chat Model, Gemini Chat Model (PaLM) and the node Gemini.