Gemini 3.5 Flash Low: Google's New AI Reduces Token Use

Google has launched a new AI model called Gemini 3.5 Flash Low. This model aims to reduce token consumption for users of its AI coding platform, Antigravity. What's interesting is this new variant addresses concerns about token usage and rate limits. Developers reported that even simple tasks were exhausting their token quotas after Google shifted to a compute-based usage system. The 3.5 Flash Low model is designed to consume notably fewer tokens. Google states it uses about 45 percent fewer tokens in its output compared to previous versions. Despite this, the company emphasizes there is no compromise on performance, claiming superior results in software engineering tasks. Google has also restructured its Flash model lineup into three versions: Low, Medium, and High, to cater to different user needs. Additionally, Google has reset the Gemini quota for all plans this week, including for free users. This helps users explore new AI tools without immediate restrictions.