According to CoinMeta, three weeks after the launch of Google's Gemini 3.7 Flash, the intelligence score increased by another 4 points to reach 56 points. The new model continues to focus on programming and agent. It has already been launched on platforms such as Gemini API, AI Studio, Antigravity, and has begun to be used in Gemini Spark. Google states that in the first week of its launch, Gemini 3.7 Flash broke the previous growth record set by the Gemini model, becoming Google's fastest-growing Gemini model. The high-inference mode of this model has an output speed of about 340 token per second, with each evaluation task costing approximately $0.40, thus entering the Pareto frontier of intelligence and completion time. The cost of the medium-inference mode has been reduced to $0.26, also entering the Pareto frontier of intelligence and cost.