Google Unveils Gemini 3.7 Flash, Three Weeks After Introducing Gemini 3.6 Flash

UK August 14 ,2026 -  Google has unveiled Gemini 3.7 Flash, a new artificial intelligence model designed to improve coding, web development, and AI agent workflows.

 The company said on Friday that the model is the most intelligent model yet for coding and agents, with the release coming less than a month after the launch of Gemini 3.6 Flash.

Developers can access the model through Google's AI Studio and Android Studio, while enterprises can access it through Google's Gemini Enterprise Agent Platform and Gemini Enterprise app.

 On the other hand, individuals can access the model via Spark, the personal agent in the Gemini app for Google AI Pro and Ultra subscribers in supported countries.

The model is available at an introductory price of $0.75 per one million input tokens and $3.75 per one million output tokens through the end of 2026, allowing developers to deploy AI agents at lower costs.

“This release comes just three weeks after Gemini 3.6 Flash, and is a direct result of developer feedback and algorithmic innovations that we look forward to bringing to future models,” the company stated.

Gemini 3.7 Flash delivers key improvements in software engineering, knowledge work and web development at half the original cost of Gemini 3.6 Flash per million tokens, according to the company.

 Google said that the model recorded higher performance compared to its predecessor in several coding benchmarks.

 The model achieved 43.6 per cent accuracy on FrontierCode 1.1 Main, compared with 34.4 per cent for Gemini 3.6 Flash, while its score on DeepSWE v1.1 rose to 65.3 per cent from 49 per cent.

In web development, Google said Gemini 3.7 Flash can generate more functional layouts and feature-complete applications using fewer prompts. The model also scored higher than Gemini 3.6 Flash on WebDev Arena, recording an Elo score of 1,588 compared with 1,538.

Google said Gemini 3.7 Flash also offers developers an improved experience by adapting better to roadblocks, clarifying user intent when necessary and following instructions more accurately.

The company has further assured that the new model also includes updated safeguards designed to address potential misuse involving chemical, biological, radiological and nuclear threats, as well as cyber offences.