Gemini 4 Argon: Global Launch, Pricing, and Practical Uses

Google is globally rolling out Gemini 4 Argon for developers and businesses, with an introductory price of $2 per million input tokens and $10 for output. The model is touted as highly effective for multi-step tasks and ranks number one on the Vals Index. Reported use cases range from cybersecurity to codebase optimization, and even the synthesis of massive scientific data.

Cybersecurity: Fairwind Expansion and Wiz Feedback

Google is beginning to expand its Fairwind program to hundreds of additional partners. Wiz is already utilizing Argon in its Scan for Good initiative, which provides free protection for critical infrastructures. With early access, Wiz reports an impact across several high-stakes sectors. The company states that in collaboration with a federal agency, Argon helped identify and rectify a vulnerability that could have exposed sensitive government data and infrastructures, contributing to the prevention of critical information exposure and unauthorized access to specialized models. In healthcare, Wiz indicates it quickly discovered and patched a public download bypass that exposed a master key and granted administrator access, claiming proactive protection for hundreds of thousands of patient records. More broadly, Argon is presented as aiding cybersecurity by auditing critical codebases and formal proofs, and detecting zero-day vulnerabilities and logic flaws.

Software Engineering: Refactoring and Measured Accelerations

Google attributes recent software productivity gains to Argon, with a more than 35% increase in agent code submissions and a roughly one-third decrease in returns. Regarding Chrome, the tool has enabled the separation of web page management from a set of browser features comprising about 6 million lines within a Chromium codebase totaling 40 million lines, facilitating an architectural overhaul involving hundreds of files in just a few weeks. According to Google, this change will allow pages to be moved between different tabs, side panels, or floating windows without needing to reload, while eliminating a certain category of crashes. In the realm of agents, optimizations now integrated into google/adk-python reduce memory preparation time by 40% during extended sessions and accelerate cold imports by 21%. For scientific computing, researchers report that xTB achieves performance up to 2.3 times better for predicting molecular structures and 1.8 times for conformer exploration. For the Gemma model family, Google reports an increase of 2.85 times in inference speed on A100 via vLLM, with throughput rising from approximately 37,900 to over 108,000 tokens per second, achieving full token parity in just a few days. Google also highlights advancements in spatial reasoning, mathematical accuracy, and visual coding, accessible via slash commands and multi-agent loops to generate 3D visualizations, games, or animations displayed in the browser.

Business and Multimodal Use Cases: From Chaco Canyon to Client Proposals

Google presents Argon as combining advanced reasoning, visual perception, and industry knowledge, with communication deemed more direct and better aligned with style guidelines than previous models. Companies and startups with early access report results. The model is highlighted for its understanding of multimodal documents, time series analysis, and interactive analytical exploration. A detailed example involves the 3D reconstruction of Chaco Canyon from about 50 GB of LiDAR scans, GIS vectors, CAD plans, and archives to create explorable digital heritage sites. In the Gemini application, Argon can transform an unstructured file into a client presentation, then automate the sending and scheduling of future requests. It is also shown synthesizing data from three open-source repositories into a 3D medical educational tool and generating 3D printable furniture pieces from a sketch and a photo. Isomorphic Labs uses Argon to accelerate drug research and design; according to Google, the rapid synthesis of PDFs and patents reduces a weeks-long workload to just a few hours.

Availability and Pricing: API, Applications, and Discount on Cached Tokens

Gemini 4 Argon is now globally accessible via the API in AI Studio and on the Gemini platform for developers and businesses. The model is also available to Google AI Ultra subscribers in Antigravity, as well as to Google AI Ultra and Pro subscribers in the Gemini application and in the AI mode of search. The launch price is set at $2 per million input tokens and $10 per million output tokens, with a 95% discount applied to cached input tokens.

Positioning and Research: Vals Index and Quantum Work

Argon is presented by Google as capable of handling complex tasks across various fields, and it ranks number one on the Vals Index, which evaluates model performance in sectors such as finance, coding, law, and taxation. Google emphasizes that internal teams, companies, and researchers are already leveraging the model. The tool has made an impact across several disciplines and has contributed to advancements in error correction and algorithms in quantum computing.