Google LLC has launched Gemini 3.7 Flash, its latest Flash-series artificial intelligence model designed to improve software development, web development, knowledge work, and AI agent workflows. The new model was introduced on August 13, just three weeks after Gemini 3.6 Flash, highlighting Google’s accelerated pace of AI model development.
According to Google, Gemini 3.7 Flash delivers substantial improvements in coding, multi-step planning, instruction following, tool use, and web development. The company describes it as its most intelligent Flash workhorse model yet for coding and agent-based applications.
One of the major improvements is in software engineering. Gemini 3.7 Flash is designed to handle complex coding tasks, including researching codebases, generating and testing code, verifying results, and working through multiple steps with fewer interruptions. The model is also optimized for full-stack development and agentic coding workflows.
Google also highlighted improvements in web and user-interface development. Gemini 3.7 Flash can generate layouts that more closely match reference images provided by users, while producing more complete applications with fewer prompts. These capabilities could make the model particularly useful for developers building websites, applications, and prototypes.
The model supports a 1 million-token context window, allowing it to process large amounts of text, images, and video within a single prompt. It can generate up to 64,000 tokens of output, making it suitable for longer and more complex development workflows.
Gemini 3.7 Flash also showed improvements in knowledge-work benchmarks. Google reported that its GDP.pdf score increased to 34%, compared with 22% for Gemini 3.6 Flash. The company also reported stronger performance on agentic and automation-focused evaluations.
Another important focus is AI agents. Gemini 3.7 Flash is designed to handle multi-step workflows, make tool calls, adapt when it encounters obstacles, and follow instructions more accurately. This makes it suitable for applications where AI systems need to complete tasks rather than simply generate text or code. Google says the model can also support agent ensembles and enterprise workflows.
Google is also making Gemini 3.7 Flash more affordable. During its introductory pricing period, the model is being offered at half the original price of Gemini 3.6 Flash. This lower cost could make it attractive for developers running AI agents that require repeated model calls, planning steps, and tool interactions.
The model is becoming available across Google’s developer ecosystem and is also rolling out to Gemini Spark, the company’s AI agent service. Gemini 3.7 Flash is additionally being integrated into developer tools such as GitHub Copilot, where it can be used for coding, web development, codebase research, and agentic programming workflows.
Google has not publicly detailed the complete architecture or training process behind Gemini 3.7 Flash. Its model card indicates that the model is intended for users, developers, and enterprises, with particular applications in coding, agentic workflows, and enterprise tasks.
With Gemini 3.7 Flash, Google is positioning its latest Flash model as a faster and more cost-efficient option for developers and businesses that need advanced coding and AI-agent capabilities. The combination of improved performance, a large context window, multimodal processing, and lower introductory pricing could make Gemini 3.7 Flash an important addition to Google’s growing AI development ecosystem.
