Regression with Canvas and 3.1 Pro in Gemini App

Since Google made their announcement (March 11) about adjusting plans and limits, using canvas+pro appears to route to a different or lesser model.

I first noticed that completions were coming extremely quick and sometimes no thinking output would be included.

I thought this was due to my limits or plan, I am on Workspace Business Standard with no add-on, but this happens even with users that have never used Gemini before. Looking over the new support pages implies nothing should have changed about my access! Even without an add on.

If canvas limits have been changed, that needs to be reflected somewhere.


Left: canvas enabled. Right: standard chat only. This user had not used Gemini before.

(I understand that this screenshot is not the most reliable signal, as it could just be a hallucination. Yet my own A/B testing spanning different users and multiple attempts shows a drastic difference in output and project comprehension.)