Google releases AI model “Gemini 3.7 Flash”, greatly improving coding and agent processing capabilities

On August 14th, Google released the latest version of its AI model, ``Gemini 3.7 Flash.'' This update came just three weeks after the release of the previous generation "Gemini 3.6 Flash", and it features significantly improved coding functions and AI agent processing power for developers. It can be used through development environments such as ``Google AI Studio'' and ``Android Studio.''

Until the end of 2026, API usage fees will be $0.75 per 1 million input tokens and $3.75 per 1 million output tokens, half the previous price.

Gemini 3.7 Flash has improved over its predecessor in benchmark tests that measure the "accuracy" of software development capabilities. Specifically, "FrontierCode 1.1 Main" increased from 34.4% to 43.6%, and "DeepSWE v1.1" increased from 49.0% to 65.3%. It is said to outperform conventional models in terms of automation performance for complex document processing and business workflows.

On the safety front, it incorporates safeguards to prevent misuse in the chemical, biological, radiological, and nuclear (CBRN) and cyber attack domains, based on a bioresilience approach and cyber program.

This model can also be used with ``Gemini Spark,'' which is available for paid plan users of Google's AI agent ``Gemini.''