Google has taken another step toward solidifying its artificial intelligence offering in the enterprise sector. The company has integrated its Antigravity agent development platform into eligible Gemini Enterprise subscriptions, while also introducing the more affordable and capable Gemini 3.7 Flash model, designed for advanced programming and automation. These two additions are part of a broader effort to position Gemini as an enterprise AI solution, capable of securely managing agents and providing centralized cost control.
The announcement, which has circulated through various specialized sources, coincides with a time when companies are looking to integrate AI into their workflows but need to maintain clear boundaries on access to data, systems, and resources. The arrival of Antigravity to Gemini Enterprise provides precisely that: management tools and security policies that allow organizations to deploy AI agents in critical environments without compromising regulatory compliance or privacy. Furthermore, the arrival of Gemini 3.7 Flash, with substantial performance improvements and a reduced price, reinforces the value proposition of the entire Gemini Enterprise ecosystem.
Antigravity is making its way into the business world
Antigravity, which began as a platform for agentic AI developers, is now part of Gemini Enterprise subscriptions. This integration allows organizations using Gemini's enterprise suite to access agent development tools with centralized management. According to the official documentation, the platform is already available as an extension for Visual Studio Code, Visual Studio, JetBrains, and Zed environments, a move that facilitates adoption without requiring developers to switch from their usual tools.
But the feature that most interests companies is control: administrators can set restrictions on the files Antigravity can access, manage terminal usage, and control which websites the agent can visit. All of this is monitored through an expense dashboard that links to the organization's consumption policies. In this way, agent usage doesn't become an uncontrolled risk, but rather can be audited and limited according to each project. "Administration is not just a technical function, but an operational requirement," the Google team points out.
Security managers at companies welcome these features, as they allow an AI agent to operate with partial permissions and with full traceability of its actions. In a context where autonomous models can modify code or execute tasks on behalf of a user, having defined limits is essential to reduce errors and potential abuse. The documentation emphasizes that consumption and access policies can be tailored to different roles within an organization, and that human oversight should remain the final recourse for critical decisions.
Gemini 3.7 Flash: a leap in quality and efficiency
Along with the release of Antigravity, Google has announced Gemini 3.7 Flash, a model that represents a substantial improvement over its predecessor, 3.6 Flash. This new model has been designed to excel in software engineering, data analysis, and web automation tasks. According to the results released by the company, in tests such as FrontierCode 1.1 and DeepSWE v1.1, Gemini 3.7 Flash shows a notable improvement over the previous version: in the first benchmark, it rises from 34,4% to 43,6%, and in the second, from 49,0% to 65,3%. These metrics reflect an improvement in the quality of the code generated on the first attempt, as well as more effective debugging.
In the area of enterprise automation, the AutomationBench test shows an improvement from 17,0% to 30,4% in the ability to manage documents and business processes. Its performance was also evaluated in specific PDF and document management tasks, where it reached 34,0% compared to 22,0% for its predecessor. The data suggests that the model is particularly well-suited for environments where reliability and productivity are needed without excessive instruction.
The launch of the model, as is typical for Google, comes with a pricing policy. The new Gemini 3.7 Flash will cost $0,75 per million tokens to enter and $3,75 per million tokens to exit. This represents a 50% reduction compared to its predecessor. The company has confirmed that this promotional price will be available until the end of the year, with the goal of enabling teams to scale their AI applications cost-effectively. Google makes it clear that this policy aims to "democratize access to advanced AI," something that is crucial in a corporate environment.
Security and governance for the future of enterprise AI
The adoption of these types of tools also presents security challenges. Google has integrated safeguards into Gemini 3.7 Flash to prevent its use in chemical, biological, radiological, and nuclear (CBRN) environments, as well as to prevent cyberattacks. These protective measures apply to both individual users and organizations, as described in official reports. The intention is to generate greater confidence in the corporate environment, where the risks can be greater than those assumed at a personal level.
For businesses, governance extends beyond technical controls. Organizations will need to align the use of these tools with their own internal policies regarding permissions, monitoring, and compliance. The advantage of Gemini Enterprise is that it centralizes administration, budgeting, and access functions in a single environment, although this doesn't replace the need for each company to define its own review policies. The goal is for AI agents to operate within a regulated framework without creating an additional burden for security teams.
Furthermore, the integration of Antigravity and the new Flash model isn't limited to those already mentioned. Google has confirmed that Gemini 3.7 Flash will be available on platforms such as Google Antigravity, AI Studio, Android Studio, Gemini Enterprise Agent Platform, and the Gemini Enterprise application itself. The expansion is extensive, and the company has indicated that it has taken into account suggestions from the developer community to fine-tune the tool. The published performance data supports the idea that this is a model optimized for teamwork and the automation of critical tasks.
By combining a business agent platform with a more powerful and cost-effective AI model, Google is setting a new standard for managing AI in organizations. It's not just about having the technology, but about being able to control, budget, and tailor it to each specific need. In a landscape of increasingly ambitious automation, this approach to balanced governance will help companies feel confident in embracing advanced artificial intelligence. Ultimately, the future lies in tools that are not only capable of execution, but also of being managed and audited.