Google launches Gemini 4 Argon with use cases and introductory pricing

Google is rolling out Gemini 4 Argon globally through its development interfaces and paid consumer offerings. The model is positioned for demanding multi-step tasks and comes with an introductory pricing. Internal teams, security partners, and life sciences players are already reporting measurable effects on real workloads.

Cybersecurity: Wiz assigns data and health protections to Argon

Wiz is utilizing Argon in its Scan for Good initiative, which aims to protect critical infrastructures for free. By gaining early access to the model, Wiz reports observed impacts across various sectors deemed critical. During a cybersecurity resilience program conducted with a federal agency, Argon was used to proactively identify and remediate a vulnerability that could compromise sensitive data and government infrastructures, preventing the leak of information regarding advancements in security and unauthorized access to specialized models. Wiz also reports having identified and promptly fixed a bypass vulnerability for public file uploads in the health sector, which exposed a master key and granted administrator privileges, thereby ensuring the proactive protection of hundreds of thousands of private patient records. Meanwhile, Google has announced that the Fairwind program will be expanded starting today to several hundred additional partners.

Software Engineering: Chrome Refactoring and Measured Accelerations

For Chromium, Argon has helped the team extract web page management from a legacy directory containing around six million lines, within a codebase totaling nearly forty million lines. By concurrently executing several code cleanup operations, it was possible to achieve an architectural refactoring involving hundreds of files in just a few weeks, whereas such tasks previously took several months. Google anticipates that for Chrome users, live web pages will be able to move between tabs, side panels, and floating windows without reloading, and claims that an entire category of browser crashes will be eliminated. In development workflows, Google also reports a more than 35% increase in agent code submissions and a nearly one-third reduction in code returns with Argon in Antigravity.

Optimizations of Frameworks and Scientific Libraries

For the Agent Development Kit, the team provided Argon with the .agents/skills rule and a succinct goal to reduce overhead without breaking existing applications. After a night of autonomous work, the optimizations deployed in google/adk-python reduced session preparation time in memory by 40% for long sessions and accelerated cold imports by 21%. In the realm of scientific computing, researchers have used Antigravity with Argon to enhance xTB, an open-source quantum chemistry package. The claimed contributions bring molecular structure predictions up to 2.3 times faster and conformer searches 1.8 times faster, with changes deposited on the official xTB GitHub.

Inference Throughput and Visual Generation Capabilities

Google reports a 2.85-fold increase in end-to-end inference speed thanks to vLLM on GPU A100 for Gemma 4 E2B, with throughput rising from about 37,900 to over 108,000 tokens per second, all while maintaining total token parity and without degrading quality. According to the company, this level of performance was achieved in just a few days, whereas such progress typically requires several months. Compared to previous Gemini models, Argon is presented with improved spatial reasoning, enhanced mathematical accuracy, and strengthened visual coding. Teams are using it, via commands like /goal and multi-agent loops with custom skills in Antigravity, to prototype 3D visualizations, high-fidelity games, and interactive animations rendered directly in the browser.

Business Workflows: Multimodal Documents, Heritage 3D, and Automation

Argon is described by Google as combining advanced reasoning, deep visual perception, domain knowledge, and improved writing in law, finance, science, and programming. The company highlights more direct communication, reduced hesitation, and better style adherence compared to previous Gemini models, with positive feedback from early access customers. On documents and data, Argon is presented as effective for analyzing multimodal content and longitudinal series, as well as for analytical sandboxes. Among the demonstrations, Argon synthesizes about 50 GB of airborne LiDAR, GIS vectors, CAD drawings from the Library of Congress, and NPS archives to reconstruct 3D models of Chaco Canyon, transforming fragmented archives into explorable digital sites. In the Gemini application, the model can format an unstructured project folder into a client proposal, send it, and create an automated flow for upcoming requests. It can also aggregate data from three open-source repositories for a 3D medical learning tool. On macOS with Antigravity, it generates variations of sofa legs with Nano Banana and a printable file, based on a dimensional sketch and a photo to produce a model to the correct size.

Sciences and Health: From PDFs to Quantum Algorithms

Isomorphic Labs reports that Argon is being used as a partner to enhance its drug research and development processes. The model processes complex PDFs and patents in a reduced time frame to extract useful information, which, according to the company, reduces a task that previously took several weeks to just a few hours. On the internal research side, Google notes an effect of Argon on accelerating advancements in quantum error correction and algorithms, claiming a real impact already felt by scientists and experts across multiple disciplines.

Availability and Pricing: API, Antigravity, and Paid Consumer Offers

Argon is accessible via the API in AI Studio and the Gemini platform, integrated into Google Antigravity for AI Ultra subscribers, and offered in the Gemini app and AI mode of search for AI Ultra and Pro subscribers. Google credits it with a number one ranking in the Vals Index. The launch comes with an introductory price of $2 per million tokens in input and $10 per million in output, with a 95% discount on cached input tokens. Google highlights the use of Argon by its teams, businesses, and researchers, with cases in engineering, optimization of legal, financial, and scientific processes, cybersecurity, and breakthroughs in quantum computing, fusion, and life sciences. Internally, the company reports faster code delivery in recent weeks.