Google has once again made a move in its artificial intelligence strategy with the release of Gemini 3.7 Flash, an update that arrives just three weeks after Gemini 3.6 Flash. The company seeks to solidify its range of efficient models—those that aren't the most powerful in its catalog but are essential for everyday tasks such as programming, automating processes, or working with large documents. The new version launches with a promotional price that makes it a very attractive option for developers and businesses, precisely at a time when competition with OpenAI and Anthropic is intensifying.
Google's move is no coincidence. While rumors about delays to Gemini 3.5 Pro persist, the company is focused on accelerating the pace of its Flash models, which are the ones that truly drive market volume. The new version promises substantial improvements in code, agents, and document processing , and it does so with a cost per token that will be halved over the next few months. This combination, on paper, makes Gemini 3.7 Flash one of the most well-rounded options on the market today.
Technical improvements that make all the difference
Data published by Google shows a clear improvement over the previous generation. In the FrontierCode 1.1 Main test, which measures the quality of production-ready code, Gemini 3.7 Flash achieves 43,6% compared to 34,4% for Gemini 3.6 Flash. In DeepSWE v1.1, an evaluation focused on long-term software engineering tasks, the score rises to 65,3%, surpassing the 49% of its predecessor. These figures indicate that the model generates more reliable code on the first try , which can translate into fewer iterations and, therefore, real savings in time and money for development teams.
Web development also benefits. In Arena.ai's WebDev Arena, Gemini 3.7 Flash achieved an Elo rating of 1588, up from 1538 for the previous version. The company claims the model can generate complete applications and functional mockups with fewer instructions, and can even replicate interface styles from screenshots or design systems. This ability to understand visual references opens the door to more agile workflows for designers and developers working closely together.
More autonomy for AI agents
One of the most outstanding aspects of Gemini 3.7 Flash is its performance in tasks requiring planning and the use of tools. In the AutomationBench benchmark, which evaluates the automation of business workflows, the model achieves 30,4% effectiveness, almost double the 17% recorded by Gemini 3.6 Flash. This means the model is capable of completing complex tasks with less human intervention , a key factor for companies looking to reduce operating costs.
In the area of ​​processing complex documents, the improvement is remarkable. In the GDP.pdf test, which measures information extraction from dense documents, Gemini 3.7 Flash scores 34%, compared to 22% for its predecessor. This advancement is particularly relevant for sectors such as finance, law, and life sciences , where accurately handling large volumes of documentation is a daily necessity. Google emphasizes that the model dedicates more effort to multi-step planning and tool calls, which reduces errors in long workflows.
Price and availability: the great selling point
Google's pricing strategy is undoubtedly one of the most striking aspects of this launch. Until December 31, 2026, Gemini 3.7 Flash will cost $0,75 per million entry tokens and $3,75 per million exit tokens, including reasoning tokens. This is half the original price of Gemini 3.6 Flash , a reduction the company presents as an opportunity for developers to scale their agents profitably. Starting January 1, 2027, the fees will increase to $1,50 and $7,50 respectively, making the coming months an ideal window to test the model.
The model is now available in the Gemini API, Google AI Studio, Android Studio, Google Antigravity, and the Gemini Enterprise platform. It has also been integrated into Gemini Spark, the personal AI agent available to AI Pro and Ultra subscribers in over 160 countries. This integration allows Spark to be more efficient in tasks such as consolidating files, composing emails, and updating status documents, all powered by Gemini 3.7 Flash.
Security and context in the AI ​​race
Google has also emphasized that security remains a top priority. Gemini Flash 3.7 incorporates updated safeguards to prevent misuse in areas such as chemical, biological, radiological, and nuclear (CBRN) applications, as well as cyberattacks. The company asserts that these protections do not hinder legitimate use cases , a crucial balance for maintaining the trust of developers and businesses.
The launch of Gemini 3.7 Flash comes at a time of uncertainty for Google's roadmap. Delays in Gemini 3.5 Pro have raised doubts among investors, who are wondering if the company can keep pace with its competitors. Meanwhile, the Flash range is becoming Google's commercial workhorse , with increasingly shorter update cycles and aggressive pricing. The recent reorganization at Google DeepMind, with Demis Hassabis as president and Koray Kavukcuoglu as director, reinforces the idea that the company wants to accelerate the development of its models and their commercialization.
For technology managers, the decision to adopt Gemini 3.7 Flash will depend on more than just benchmarks. The reduced price during 2026 is a clear incentive, but the true cost per completed task, including retries and human oversight, will need to be measured. The discount window allows for testing the model with low financial risk , but implementations designed now must take into account that the price will double in January. The key will be evaluating whether the model's performance justifies the long-term investment, something that only time and real-world use cases will be able to confirm.