Google Restructures Gemini Models to Handle Usage Pressure..

Google has reportedly restructured its Gemini usage system into a compute-based token model, which led to heavy complaints from users of its agentic coding platform Antigravity. After the shift from message-based limits to token-based consumption, users began hitting rate limits quickly, forcing Google to increase quotas twice. However, demand still remained high, creating performance and usage concerns across plans, including the free tier.

To address this, Google DeepMind has introduced a new variant called Gemini 3.5 Flash (Low), designed to reduce token consumption for simpler tasks without affecting performance. The original model has now been rebranded as Gemini 3.5 Flash (Medium), while a High version is aimed at complex tasks. The Low variant reportedly uses about 45% fewer tokens while maintaining strong coding performance, as Google also reset quotas across plans to ease user pressure.