Google DeepMind Debuts Gemini 3.7 Flash at Half Price

Google DeepMind has released Gemini 3.7 Flash, the latest iteration of its lightweight model series, arriving a mere three weeks after the launch of Gemini 3.6 Flash. This rapid release cycle highlights Google's aggressive development pace. The new model is immediately available in Google AI Studio, Antigravity, Android Studio, and the Gemini app. To accelerate adoption, Google is offering introductory pricing through the end of 2026 at $0.75 per million input tokens and $3.75 per million output tokens, which represents a 50 percent discount compared to the cost of Gemini 3.6 Flash.

The update delivers substantial performance improvements, particularly in software engineering and web development. On the DeepSWE v1.1 benchmark, which measures a model's ability to autonomously resolve real-world software issues, Gemini 3.7 Flash scored 65.3 percent, up from the 49.0 percent achieved by Gemini 3.6 Flash. It also improved on the FrontierCode 1.1 Main benchmark, rising to 43.6 percent from 34.4 percent. For web developers, the model scored 1588 on Arena.ai's WebDev Arena, outperforming its predecessor's Elo score of 1538, meaning it can generate functional layouts and complete applications with fewer prompts.

Beyond coding, Gemini 3.7 Flash nearly doubles performance on knowledge work tasks. Its score on AutomationBench jumped to 30.4 percent, compared to just 17.0 percent for Gemini 3.6 Flash. Additionally, Google has upgraded Gemini Spark, the 24/7 personal agent for Pro and Ultra subscribers, to run on the new model, enhancing its integration and tool use within Google Workspace. For practitioners, these changes mean they can deploy highly capable agents and automated coding workflows at a fraction of the previous cost, making complex agentic workflows economically viable for the first time.