Technology Desk: The use of Artificial Intelligence i.e. AI is continuously increasing. From studies, content creation to coding and business work, AI models are now helping in a variety of tasks. Meanwhile, Google has introduced the new Gemini 3.7 Flash in the Flash series of its Gemini model. Google has introduced it as its most capable “workhorse” model, which is specially designed for coding, software engineering, web development and AI agents.
The special thing is that Gemini 3.7 Flash has been introduced about three weeks after the launch of Gemini 3.6 Flash. Gemini 3.6 Flash was launched by Google on July 21, 2026. That model was also designed keeping in mind better speed, lower cost and more reliable performance for AI agents.
What is special in Gemini 3.7 Flash?
The main focus of Gemini 3.7 Flash is not just to answer questions. It is designed for tasks where AI has to work through multiple steps to accomplish a goal. For example, a simple chatbot can provide code if someone asks to build a website. But an AI agent can be designed in such a way that it understands the need, plans the work, writes the code, uses the necessary tools, checks the results and changes its approach when problems arise.
Gemini 3.7 Flash focuses more on similar agentic workflows. According to Google, the model has brought improvements in understanding complex instructions, planning multiple steps, and using tools.
more work with less instructions
A major feature of the new model is said to be better instruction following. This means developers may be less likely to need to give individual instructions to the AI about each small step. If there is an obstacle in the middle of any work, the model can understand the situation and change its approach. It is also capable of asking for additional information from the user if needed. This feature is important for AI agents, because in long and complex tasks, human intervention at every step can slow down the process.
Big improvement in coding
In Gemini 3.7 Flash, most of the focus has been given to coding and software engineering. According to Google data, Gemini 3.7 Flash has achieved a score of 43.6 percent in the FrontierCode 1.1 Main test. The score of Gemini 3.6 Flash in this test was 34.4 percent. In this way, there is a difference of about 9.2 percentage points between the two models.
The new model has also recorded better performance than the old model in other coding tests like DeepSWE v1.1. As per the available reports, the Gemini 3.7 Flash scored around 65.3 percent, while the Gemini 3.6 Flash stood at around 49 percent. It can be used for writing code as well as for longer and more complex software engineering tasks.
Also beneficial in creating website and UI
Gemini 3.7 The use of Flash is not limited to programming only. The model has also been improved for web development and interface creation. Developers can use it to create a similar interface by providing a screenshot, reference image or design system as input.
Here the work of the model is not only to write HTML or CSS, but also to understand the given design and prepare the interface accordingly. Google has also emphasized on improving design adherence in the new model, i.e. the ability to produce output as per the given design and instructions.
Why is AI useful for agents?
AI agents work a little differently from normal chatbots. A chatbot typically answers a user's question, while an agent may take several different steps to accomplish a goal. Suppose a company has to prepare a report. The agent may have to perform many tasks like gathering information from different sources, organizing the data, analyzing it and preparing reports. In such tasks, the model not only needs to understand the language, but also needs the ability to plan, reason, use tools and try again when a problem arises. Gemini 3.7 Flash has been designed keeping these needs in mind.
Also used in Gemini Spark
Google's AI agent Gemini Spark has also been upgraded with Gemini 3.7 Flash. Google introduced Gemini Spark in India in July 2026. It can work with Google Workspace related tools like Gmail, Docs and Sheets. The purpose of Gemini Spark is not just to answer user questions, but to help complete digital tasks after receiving instructions. The improved planning and tool-use capabilities of the new Flash model may be useful for such agent-based tasks.
Available at a lower cost than ever before
Gemini 3.7 Flash has been made available to developers via API and on several of Google's developer platforms. At the time of launch, Google has priced it at $0.75 per 10 lakh input tokens and $3.75 per 10 lakh output tokens for a limited period. This starting price is till December 31, 2026. After this the price will increase to $1.50 and $7.50 per 1 million tokens respectively. This price also makes model cost an important consideration for developers building long-running AI agents and using APIs at scale.
Where to get Gemini 3.7 Flash?
Gemini 3.7 Flash has been made available to developers on platforms such as Gemini API, Google AI Studio, Google Antigravity, and Android Studio. For enterprise users, it can also be used on Gemini Enterprise related platforms. The Gemini Flash series is already being used by Google among developers as fast and cost-efficient AI models. Gemini 3 Flash was made available by Google in 2025 on platforms like API, AI Studio, Gemini CLI, Android Studio and Vertex AI.
Follow the LALLURAM.COM MP channel on WhatsApp