Google Launches Gemini 4 Argon for Global AI Users
Unlike models designed mainly for quick responses, Argon focuses on multi-step work that requires sustained reasoning. Moreover, Google says the model can maintain context across lengthy tasks and handle demanding workflows with greater precision. Its capabilities include analysing documents, modifying large codebases, and supporting technical research.
The launch strengthens Google’s effort to bring advanced AI into professional workflows. Consequently, businesses can use Argon to explore complex problems, automate selected processes, and build more capable AI agents through Google’s developer and enterprise platforms.
Gemini 4 Argon Targets Complex Coding and Research
Google says Argon delivers strong performance across several economically important fields. According to the company, it ranks first on the Vals Index, which evaluates AI models across areas such as finance, coding, legal work, and tax tasks. However, these results reflect a company-reported evaluation and should be considered alongside independent testing.
The model also supports a range of input formats. Developers can submit text, images, audio, video, and PDF files, and the combined input-output context limit is approximately 1 million tokens. As a result, Argon can work with substantial amounts of information during complex tasks.
Google has also reported gains from using Argon within its Antigravity development platform. The company says agent-driven code submissions increased by more than 35%, while code rollbacks fell by nearly one-third. In addition, Argon helped engineers reorganise parts of Chromium’s large codebase, completing work that Google says would otherwise have taken months in weeks.
Beyond software development, Google identifies applications in financial analysis, legal workflows, cybersecurity defence, and scientific research. For example, Argon has supported work involving quantum computing, nuclear fusion, and life sciences. Therefore, the company is positioning the model for organisations that need AI to handle demanding, specialised work.
Global Availability, Pricing and Safety Measures
Gemini 4 Argon is available globally through the Gemini API in Google AI Studio and the Gemini platform. Enterprise customers can also access it through Gemini for business and Google Antigravity. Meanwhile, eligible Google AI Pro and Ultra subscribers can use Argon through the Gemini app and AI Mode in Google Search.
Google has introduced the model at the following API rates.
-
Input tokens: $2 per million tokens.
-
Output tokens: $10 per million tokens.
-
Cached input tokens: 95% discount.
These introductory rates apply through November 30, 2026. After that date, Google plans to charge $4 per million input tokens and $20 per million output tokens.
Google is also expanding access for scientific researchers and cybersecurity professionals. In particular, the company plans to offer six months of free access to Antigravity’s highest usage tier for up to 10,000 eligible researchers, subject to selection and capacity. Additionally, its Fairwind programme gives vetted cybersecurity defenders access to a version of Argon without standard cyber safeguards.
Safety remains important as Google expands the model’s capabilities. The company has tested Argon against prompt-injection attacks and reports a 0.7% attack success rate on the Grey Swan Indirect Prompt Injection benchmark under its stated testing conditions. Nevertheless, organisations should continue to assess model behaviour, access controls, and data protection before deploying it in sensitive workflows.
Ultimately, Gemini 4 Argon represents Google’s push to make advanced AI useful for sustained professional work. Its commercial impact will depend on performance in real-world deployments, API costs, and the reliability of the agents built around it.
