Google Expands Supply of 'Gemini 4 Argon' to Target Enterprise AI and Coding Markets

Google has begun broad rollout of its next-generation AI model, "Gemini 4 Argon," which it had first unveiled to a limited set of users, by expanding access to developers and enterprise customers. On the 13th (local time), Google announced the general availability of Gemini 4 Argon for developers and enterprise customers. The model, which had previously been unveiled to some users, will now be expanded to the Gemini API, Google AI Studio, and enterprise AI services.

The expanded rollout is part of Google's push into the enterprise AI market and its move into the software development market. Gemini 4 Argon was developed for complex tasks in fields that require specialized knowledge, including coding, finance, law, and science. Google emphasized the model's reasoning capabilities for step-by-step handling of complex tasks and its code-writing functions using AI agents.

Google cited strengthened multi-step problem analysis and reasoning, including result verification, as competitive advantages of Gemini 4 Argon. It also highlighted modernization of legacy systems as another key strength. Through these capabilities, Google underscored that the model is focused on tasks that require expertise and on coding support.

In terms of performance, Google said Gemini 4 Argon ranked No. 1 on the Vals Index. Vals Index is a benchmark that evaluates the ability to perform specialized work in finance, law, coding, and tax. Google used the index to emphasize Gemini 4 Argon's ability to handle specialized tasks.

Along with the expanded rollout, Google has integrated Argon into Google Antigravity, its AI agent-based development platform. As a result, developers can now convey work goals through natural-language instructions. Argon analyzes existing code and also handles new feature development, bug fixes, and testing.

Before the general release, Google validated Argon's performance within its internal engineering organization. The testing period lasted about 4 weeks. As a result, the share of code written by AI agents increased by 35%, while the rollback rate for deployed code fell by 30%. More than 100,000 Google employees are now using the internal version of Antigravity every month.

Argon is also being used for large-scale software architecture improvements. Google's Chrome development team applied Argon to work involving about 40 million lines of Chromium code. Of that codebase, 6 million lines were related to existing browser features. The work involved separating core webpage processing functions.

Google said it cut development time by deploying Argon for large-scale code changes and refactoring. Argon supports code changes at the level of hundreds of files, and Google said work that would have taken months under the traditional approach was completed in just weeks. As an example of AI use, Google pointed to refactoring performed while preserving functionality.

Google also said Argon's performance improvements were confirmed in ADK, its open-source AI agent development tool. Argon analyzed and optimized existing code, cutting long-session preparation time by 40% and improving initial module loading speed by 21%. The related changes were reflected in an actual open-source project.

Argon was also applied to efforts to move legacy programming-language-based systems to new development environments. Google's R&D team used Argon to convert Fortran research code to Rust, and Google said program execution performance improved by 1.8 times after the transition.

Google expanded the use of AI beyond software development to general enterprise work. The company said it is using Antigravity in its in-house sales, finance, and business operations organizations, where it is being used for data analysis, report writing, and workflow automation. It also said the approach links internal enterprise data and business systems with AI agents through plugins and MCP.

Google is providing Argon in the Gemini app for general users. For Google AI Pro and Ultra subscribers, it is offering a "Tasks" mode, through which users can request multi-step work and have it run in the background. It also supports integration with more than 40 external apps, including Instacart, Adobe, and Shopify.

Google is running a discount period to mark the launch of the API until November 30. During the discount period, API pricing is USD 2 per 1 million input tokens and USD 10 per 1 million output tokens. After the discount ends, the general-availability model will be priced at USD 4 for input and USD 20 for output per 1 million tokens.

Google plans to expand supply of Gemini 4 Argon so it can be applied to enterprises' existing development environments and business systems. Accordingly, the company is looking to broaden the use of AI agents beyond code writing and testing to include legacy system architecture improvements and workflow automation.

Source: TECHWORLD · Kim Seung-gi
Original: https://www.epnc.co.kr/news/articleView.html?idxno=408168

References

• https://www.epnc.co.kr/news/articleView.html?idxno=408168