Google launches Gemini 3.7 Flash for coding and AI agents

Google has launched Gemini 3.7 Flash, a new AI model optimized for coding, web development, and autonomous AI agents. Released only three weeks after Gemini 3.6 Flash, the update focuses on improving multi-step planning, debugging, and software engineering capabilities. Internal benchmarks show significant improvements, such as a jump from 49% to 65.3% on the DeepSWE v1.1 test.

To encourage adoption, Google is offering the model at a reduced introductory price—roughly half the cost of its predecessor—through the end of 2026. The model is being integrated into various services, including Google Search (for AI Pro and Ultra subscribers) and the Gemini Spark personal agent.

Separately, Google announced a major user milestone, with the Gemini app surpassing 1 billion monthly active users. CEO Sundar Pichai noted this as the company’s fastest-growing product ever, highlighting strong engagement through voice features and multimodal interactions. This growth comes amid a leadership reshuffle at Google DeepMind, where Demis Hassabis has transitioned to Chief Scientist role.