Today, Google DeepMind introduces Gemini 3.7 Flash, the most intelligent workhorse model yet for coding and agents. Released just three weeks after 3.6 Flash, it reflects developer feedback and algorithmic innovations. 3.7 Flash delivers substantial improvements across software engineering, knowledge work, and web development — with an introductory price of half the original 3.6 Flash cost per million tokens.

In coding, 3.7 Flash achieves higher first-pass code accuracy and stronger gains in debugging and issue resolution. It scores 43.6% vs 34.4% on Frontier Code 1.1 Main and 65.3% vs 49.0% on Deep SWE v1.1. For web development, it generates more functional layouts and outperforms 3.6 Flash on Arena.ai’s Web Dev Arena with an Elo score of 1588 vs 1538. In knowledge-dense fields, it significantly outperforms on GDP.pdf (34.0% vs 22.0%) and Automation Bench (30.4% vs 17.0%).

The developer experience is noticeably improved, with better instruction following, multi-step planning, and tool calls. 3.7 Flash is available through year-end at $0.75/1M input tokens and $3.75/1M output tokens, enabling cost-effective scaling of production-ready agents. Gemini Spark (for Pro and Ultra subscribers) will also use 3.7 Flash starting today, making the personal AI agent more efficient for knowledge work and multi-skill workflows.

Safety safeguards have been updated for CBRN and cyber offense. Developers can start building via Google AI Studio and Android Studio; enterprises access via Gemini Enterprise Agent Platform; individuals via the Gemini app in supported countries.