TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Google announced Gemini 4 Argon on Oct. 13, 2026, making the model available through its API, Gemini app, Search AI Mode and Antigravity, with access varying by user type. Google says Argon is designed for complex, multistep work and cites internal engineering results, but independent evaluations and several deployment details are not provided in the announcement.
Google announced Gemini 4 Argon on Oct. 13, saying the model is available from launch to developers and enterprises through its API and to some paid subscribers through Google products. The company is positioning Argon for complex, multistep work in areas such as coding, finance, law, cybersecurity and science, and says its launch price is $2 per million input tokens and $10 per million output tokens.
According to Google, developers and enterprises can access Argon through the API in AI Studio and the Gemini platform. Google AI Ultra subscribers can use it in Google Antigravity, while AI Ultra and Pro subscribers can access it in the Gemini app and AI Mode in Search. The announcement does not describe access for free-tier users or provide a full account of regional availability beyond calling the release global.
Google says Argon performs well on complex tasks across several professional fields and ranks No. 1 on the Vals Index, which the company describes as measuring model performance in economic sectors including finance, coding, legal and tax work. The post does not supply the underlying scores, test methodology, comparison set or independent verification. The ranking should therefore be understood as a claim reported by Google, not as a separately established assessment.
The company also reports results from its own engineering use. It says Argon-assisted work in Antigravity increased agentic code submissions by more than 35% and reduced code rollbacks by nearly a third. Other examples include a Chrome code refactor, optimizations to Google’s open-source Agent Development Kit, and speed improvements to the xTB quantum chemistry package. These are examples and measurements supplied by Google; the announcement does not provide independent replication or a common test framework for comparing them.
Argon Targets High-Stakes Work
Argon’s launch matters because Google is presenting it not simply as a general-purpose chatbot, but as a model for work involving long task sequences, large codebases and technical documents. If its capabilities and Google’s reported results hold up in outside testing, the model could affect how organizations approach software maintenance, document analysis and scientific computing. The practical value will depend on accuracy, reliability, integration costs and how much review human experts still need to provide.
The advertised token price gives developers an initial way to estimate usage costs: $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95%, according to Google. Those rates do not by themselves establish the total cost of using Argon in a product. The announcement does not specify typical task sizes, service limits or how pricing may change after the introductory period.
For businesses and public-facing services, claims about legal, financial, cybersecurity and scientific performance call for particular care. Google’s announcement describes potential uses and internal examples, but does not establish that Argon can make high-stakes decisions without expert review. Buyers will need to assess it against their own requirements, including data handling, error rates and the consequences of incorrect outputs.
As an affiliate, we earn on qualifying purchases.
How Google Describes Argon
Google’s Oct. 13 post presents Argon as its latest frontier model and says it is intended for developers, enterprises, researchers and consumers. The release links access to existing Google products and subscription tiers, rather than describing a single universal route for all users. Its stated use cases span software development, professional workflows, research and cybersecurity.
The announcement gives several engineering examples to support Google’s description. It says Argon helped reorganize browser feature code within Chromium, a codebase the post describes as containing about 40 million lines; the legacy folder discussed in that example contains roughly six million lines. Google says the work was completed across hundreds of files and could support moving live webpages among browser tabs, side panels and floating windows without reloading. That future browser behavior is described as an anticipated benefit, not a feature the post says is already available to users.
Other reported measurements concern different tasks and conditions. Google says its Agent Development Kit changes reduced in-memory session preparation time by 40% in long sessions and improved cold imports by 21%. It reports that xTB became up to 2.3 times faster for molecular structure prediction and 1.8 times faster for conformer search. These figures are specific to the examples in the post and should not be treated as general performance guarantees for Argon.
“Gemini 4 Argon is now globally available today”
— Google, in its Gemini 4 Argon announcement
As an affiliate, we earn on qualifying purchases.
Independent Testing Still Needed
Google’s post does not provide benchmark scores, test protocols or a list of competing models behind the Vals Index ranking. It also does not include independent evaluations showing how Argon performs on the tasks described, or how often it makes errors. The reported engineering improvements come from examples selected and described by Google, so their wider applicability remains unknown.
The announcement does not spell out Argon’s data retention and privacy terms, service limits, supported regions, or the full availability schedule for each product. It is also unclear whether the introductory token prices are temporary and, if so, when they may change. Google describes cybersecurity and other high-stakes applications, but the post does not detail safeguards, approval processes or the level of human oversight needed for those uses.
Google says Argon has been used by teams at the company, enterprises and in research, but it does not name all participating organizations or provide external customer results in the supplied material. The reported improvements should not be read as proof that every user or deployment will see similar outcomes.
AI-powered legal research software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Access and Evaluation Ahead
Developers and enterprise teams can begin testing Argon through the Gemini API in AI Studio and the Gemini platform, according to Google. Eligible subscribers can try it in Antigravity, the Gemini app and Search AI Mode, depending on their plan. Organizations considering it for professional workflows will need to check the relevant product terms and assess its performance on their own tasks before relying on it.
The next useful evidence will include independent benchmark results, detailed testing methods and reports from customers outside Google’s own teams. More information about pricing after launch, regional coverage, data policies and safety controls would help developers and businesses judge whether Argon fits their needs. Google has not specified a date for such follow-up material in the announcement.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is Gemini 4 Argon?
Gemini 4 Argon is a new Google model that the company says is intended for complex tasks, including coding, finance, legal work, cybersecurity and science.
Who can access Argon?
Google says developers and enterprises can use it through the API in AI Studio and the Gemini platform. AI Ultra subscribers can access it in Antigravity; AI Ultra and Pro subscribers can use it in the Gemini app and AI Mode in Search.
How much does Argon cost?
Google lists an introductory price of $2 per million input tokens and $10 per million output tokens. Cached input tokens are discounted by 95%. The announcement does not say when or whether these prices will change.
Has Argon’s performance been independently confirmed?
The announcement does not include independent test results. Its benchmark ranking and engineering improvements are claims and examples reported by Google, and the post does not provide complete methods or external verification.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
