When using the Google AI Pro subscription, I feel that the Pro subscription is too limiting. The removal of access to Gemini 3 Flash has made it much harder to get anything done now, especially due to the new Gemini 3.5 Flash model and its heightened pricing. The cost-to-performance ratio is moderately imbalanced, especially as, even in low-thinking settings, the model uses many tokens, which also impacts its pricing.
I believe it would be much better to at least keep the legacy models until they are fully sunset on the API level, allowing users like me on the Pro subscription to get more value from it. Regarding credit usage, Gemini 3.5 Flash, regardless of the thinking option, exhausts AI credits astronomically more than Claude Opus does for the exact same task, based on my tests and monitoring. I believe that reinstating legacy options is the stronger choice, especially with more quota, and I recommend the implementation of Gemma4 models within Antigravity.