Alphabet’s Google has announced global availability of its Gemini 4 Argon AI model for developers and businesses.
The Gemini 4 Argon model is being delivered to users via the API in AI Studio and the Gemini platform.
Go deeper with GlobalData
Access deeper industry intelligence
Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.
Access is also granted to Google AI Ultra subscribers in Google Antigravity, and to Ultra and Pro subscribers in the Gemini app and AI Mode in Search.
According to Google, Argon is suited for complex tasks in coding, law, finance, cybersecurity, and science, achieving high scores in benchmarks across these sectors.
The model is ranked first on the Vals Index, a benchmark measuring model performance in economic sectors such as finance, legal, coding, and tax work.
Google reports Argon has already been implemented internally for engineering problems, workflow optimisation in finance and science, cybersecurity vulnerability detection, and scientific research.
In coding applications, teams using Argon in Antigravity reported a 35% increase in code submissions and a reduction in code rollbacks by nearly a third.
Specific examples include helping the Chrome team restructure a 40 million-line codebase, optimising the Agent Development Kit’s performance, and improving the runtime of the quantum chemistry package xTB.
The company stated it achieved a 2.85-fold speedup for its Gemma models and facilitated the development of complex 3D visualisations and visual coding in browsers.
Enterprise and startup customers participating in early access programmes have reported positive results in document comprehension, data analysis, and analytical workflows.
In cybersecurity, Google said partners such as Wiz have used Argon to identify gaps in governmental and healthcare systems, helping to prevent data exposure and patch vulnerabilities.
In scientific applications, Argon was used in quantum computing to enhance algorithm design, support research in nuclear fusion, aid rare disease diagnosis, and accelerate enzyme engineering processes.
Argon offers a one million token limit for unified input and output via the Gemini API.
Access is available globally across developer, enterprise, cybersecurity, and research segments, with some programmes requiring application or specific subscriptions.
In April 2026, Google introduced Gemma 4, a set of open AI models intended for use on various hardware, from mobile phones to personal computers.
