Google Launches Gemini 3.7 Flash with Enhanced Coding and Automation at Half Price

Google released the Gemini 3.7 Flash model on August 16, 2026, its flagship Flash-series model for coding and agent scenarios. The model arrives just three weeks after Gemini 3.6 Flash and delivers improvements in programming, web development, complex document processing, and enterprise automation. In the DeepSWE v1.1 benchmark, its score rose from 49.0% to 65.3%, close to Claude Sonnet 5 and GPT-5.6 Terra; the AutomationBench score increased from 17.0% to 30.4%.

In the Intelligence Index from third-party evaluator Artificial Analysis, Gemini 3.7 Flash (high) scored 56, similar to GPT-5.6 Terra (max), GPT-5.5 (xhigh), Grok 4.5 (high), and Claude Sonnet 5 (max). For coding, it achieved 43.6% on the FrontierCode 1.1 Main test, ahead of Gemini 3.6 Flash's 34.4%; on DeepSWE v1.1 it scored 65.3%, slightly below GPT-5.6 Terra's 69.6%. In web development, it recorded an Elo score of 1588 on the WebDev Arena test, leading other models. In knowledge-intensive tasks, it scored 34.0% on the GDP. pdf benchmark, up from the previous generation's 22.0%.

On pricing, until the end of 2026, Gemini 3.7 Flash charges $0.75 per million input tokens (approximately RMB 5.06) and $3.75 per million output tokens (approximately RMB 25.28), both half the original Gemini 3.6 Flash prices. From January 1, 2027, these prices will revert to $1.50 (approximately RMB 10.11) and $7.50 (approximately RMB 50.55) per million tokens, respectively. Gemini 3.7 Flash will also become the new base model for Google's personal AI agent Gemini Spark, helping users execute multi-step tasks across Gmail, Google Calendar, Google Docs, and other applications.

Gemini Spark began using Gemini 3.7 Flash on the launch day, enhancing tool-calling capabilities for Google Workspace applications and focusing on improving complex workflows. Spark is available to Google AI Pro and Ultra subscribers in more than 160 countries and regions. Individual users can access it through Google AI Pro or Ultra subscriptions, developers can use it via Google Antigravity, the Gemini API, Google AI Studio, and Android Studio, and enterprise customers can connect through the Gemini Enterprise Agent Platform and the Gemini Enterprise application.

Recently, the Gemini app surpassed one billion monthly active users. Google has released two generations of Flash models in just three weeks, temporarily lowered call prices, and integrated the new model into Gemini Spark. Last week, Google's AI team underwent personnel changes: Demis Hassabis stepped down as CEO of Google DeepMind to become chairman, and four core AI personnel, including Jeff Dean, left to start their own ventures.