Google Resets Gemini Limits and Releases Updated 3.5 Flash Version

Google has conducted a large-scale reset of quotas for all Gemini users, including both free and paid tiers. This decision aligns with the launch of an updated Gemini 3.5 Flash model, designed to resolve issues with sharp declines in generation quality when performing more complex tasks, particularly in software development.

Previously, Google introduced a "lightweight" version of the model—Gemini 3.5 Flash (Low)—which reduced token consumption by 45% compared to the base version. However, this led to a drop in output quality when handling tasks requiring deeper analysis. The new version aims to fix this drawback, though it remains unclear whether changes will apply only to the "Low" variant or also the "Medium" option.

The update is currently available exclusively to Antigravity users—Google’s internal platform for model testing. Similar settings are not yet available in the consumer version of Gemini. Varun Mohan, Head of Google DeepMind, confirmed the launch of the updated model, highlighting its improved stability for resource-intensive tasks.

Users are actively discussing the need for a weekly display of remaining quotas, as tracking token usage is currently impossible. Google has acknowledged this request and is considering adding such a feature in the future.