Google has introduced Gemini 3.7 Flash, a new addition to its Gemini family that the company says delivers improvements in coding, web development, reasoning, knowledge-intensive tasks and AI agent workflows.

The new model arrives only a few weeks after Gemini 3.6 Flash and is positioned as a faster and more cost-efficient option for developers and businesses building AI-powered applications. Google is also offering the model at an introductory price that is significantly lower than its predecessor.

Stronger Coding Performance

Google says Gemini 3.7 Flash has been developed to handle software engineering tasks with greater accuracy and consistency.

The model is designed to generate more production-ready code on the first attempt, potentially reducing the number of corrections and additional prompts developers need during the development process.

Its improvements are also aimed at developers working on complex software projects where following detailed instructions and managing multiple steps are important.

Better Web Development

Web development is another area where Google says Gemini 3.7 Flash has made notable progress.

The model can create more functional layouts and feature-complete web applications with fewer prompts. It is also designed to better understand visual references and maintain consistency with a given design.

Google says the model demonstrates improved design fidelity when working from screenshots, images or complete design systems.

According to the company, Gemini 3.7 Flash scored 1,588 on Arena AI's WebDev Arena, outperforming Gemini 3.6 Flash.

Improved Reasoning

Google is also targeting knowledge-intensive professional tasks with the new model.

The company says Gemini 3.7 Flash performs better on complex document processing and business workflow automation, particularly in fields such as finance, law and biosciences.

In the tests cited by Google, the model scored 34.0 per cent compared with 22.0 per cent for Gemini 3.6 Flash on GDP.pdf, while its score on AutomationBench was 30.4 per cent compared with 17.0 per cent for the earlier model.

These results are based on Google's cited evaluations and should not necessarily be treated as a guarantee of performance across every real-world use case.

Focus on AI Agents

Google has also placed greater emphasis on AI agent capabilities with Gemini 3.7 Flash.

The model is designed to follow instructions more closely, plan multi-step tasks and make tool calls more systematically. The company says this more disciplined approach could reduce the need for manual intervention and repeated attempts when AI agents are used in software engineering and other workflows.

For developers, the combination of stronger planning, coding and tool use is aimed at making the model more suitable for production-oriented AI applications.

Introductory Pricing

Google has announced an introductory pricing structure for Gemini 3.7 Flash that will remain in place through the end of 2026.

The model will cost $0.75 per one million input tokens and $3.75 per one million output tokens during the introductory period.

Google is positioning the pricing as an effort to make large-scale deployment of AI agents and production applications more affordable for developers and businesses.

New Safety Measures

The company has also introduced additional safeguards with Gemini 3.7 Flash.

Google says the model includes updated protections against potential misuse in areas involving chemical, biological, radiological and nuclear (CBRN) risks, as well as cybersecurity.

The company said it has strengthened jailbreak protections while continuing to support legitimate applications in areas covered by its broader bioresilience and cybersecurity programmes.

Model Still Has Limitations

Despite the improvements, Gemini 3.7 Flash retains some of the limitations commonly associated with foundation models.

Google notes that the model can still produce inaccurate or fabricated information, while users may occasionally encounter slow responses or timeout issues.

The company also states that the model's knowledge is current through March 2026 for some domains, while certain areas may still depend on information available only through January 2025.

Google's AI Push Continues

The launch of Gemini 3.7 Flash comes as Google continues to expand its AI model lineup and compete aggressively in coding, software development and AI agent applications.

With its focus on lower costs, stronger coding performance, multi-step reasoning and tool use, the new Flash model is aimed particularly at developers and organisations looking to integrate AI into everyday workflows.

However, as with other foundation models, users will still need to verify outputs, particularly when deploying the system for important technical, business or professional tasks.