From what I understood, Google is basically removing the old “included credits” system and moving toward a compute-based usage model similar to services like Claude, Higgsfield, and other AI platforms.
Some major changes mentioned in the mail:
• Gemini app now has compute-based limits instead of simpler model-based limits
• Usage depends on prompt complexity, thinking mode, features used, and even chat length
• Limits refresh every 5 hours, but there’s also a weekly cap
• Flash and Pro usage now seem to pull from the same overall pool more aggressively
• The included 1000 monthly AI credits for tools like Antigravity and Flow are being removed
• If you need more usage, you’ll likely have to buy additional AI credits separately
What’s interesting is that they say the “experience should remain similar,” but from actual usage recently, quota drains much faster than before — especially for heavy users doing long reasoning, image/video generation, Antigravity workflows, etc.
It honestly feels like Google is shifting toward the same controlled-usage/pay-per-compute direction as Claude and other premium AI services, except Gemini still doesn’t consistently deliver output quality or reliability at the same level in many workflows.
I especially noticed:
-
Flash no longer feels lightweight on quota
-
5-hour refreshes happen quickly
-
Heavy sessions hit limits much sooner than before
-
Antigravity usage became far more restrictive after recent updates
Curious if others are seeing the same thing or if your experience has been different after this change.