Google Unveils Gemini 4 Argon to Take On OpenAI in High-Stakes AI Work

Google has launched Gemini 4 Argon, a new artificial intelligence model designed to handle complex, multi-step tasks for businesses, developers and researchers as competition intensifies among leading AI companies.

Announced on Tuesday, October 13, the model targets demanding professional work across software development, finance, law, cybersecurity and science. Google is positioning it as a tool for solving problems that require more than generating text or answering simple questions.

The launch comes as the company seeks to strengthen its position against OpenAI and Anthropic, whose models compete for developers and businesses adopting AI for increasingly complex tasks.

Beyond Chatbots

Google says Gemini 4 Argon can work through complicated engineering problems, optimise business processes and help researchers investigate scientific questions.

The company points to its use in software development as evidence of its capabilities. According to Google, teams using Argon through its Antigravity coding platform increased submissions of AI-generated code by more than 35%, while reducing code rollbacks by nearly a third.

The model has also helped Google's engineering teams optimise software. One project involved reorganising parts of Chromium, the codebase behind its Chrome browser, with Google saying work that would normally take months was completed in weeks.

These results are based on Google's own reports of internal testing and use. Independent comparisons will be important in determining how consistently the model performs in real-world settings.

Google also says Argon ranks first on the Vals Index, a benchmark assessing AI performance across professional fields such as finance, legal work, taxation and software development.

What It Costs and Who Can Use It

Gemini 4 Argon is available worldwide to developers and businesses through Google's AI Studio and Gemini API, as well as through Google Cloud's Gemini platform.

Consumers can access it through the Gemini app and AI Mode in Google Search if they subscribe to Google AI Pro or Ultra. Access through Google's Antigravity coding tool is reserved for Ultra subscribers.

For developers using the API, introductory pricing starts at $2 per million input tokens and $10 per million output tokens. Cached input tokens receive a 95% discount. Google says the standard rates will double after November 30, 2026, to $4 per million input tokens and $20 per million output tokens.

The pricing structure gives developers a way to integrate the model into their own applications, although the final cost will depend on how much they use it and the complexity of their workloads.

Google is also offering 10,000 scientists six months of free access to the highest usage tier of Antigravity, seeking to encourage research applications.

Cybersecurity and Scientific Research

Cybersecurity is another major focus of the launch. Google plans to expand its Fairwind programme to hundreds of additional partners, providing vetted cybersecurity defenders with access to a version of Argon without its usual cyber safeguards.

The programme is intended to help security professionals identify and address vulnerabilities. However, access to a model with fewer restrictions also makes careful oversight important, given the potential for advanced AI capabilities to be misused.

Google says Argon has also supported research in areas including quantum computing, nuclear fusion and life sciences. These applications reflect the company's broader ambition to make AI useful for specialised technical work, not just everyday productivity.

The Bigger AI Race

The launch comes after a difficult period for Google's AI division. The company previously abandoned plans to release Gemini 3.5 Pro, which CEO Sundar Pichai had initially expected to launch in June. Its DeepMind organisation has also undergone leadership changes.

Gemini 4 Argon represents a fresh effort to compete in a market where AI companies are increasingly selling models on their ability to complete work, use tools and tackle complex problems rather than simply produce convincing responses.

Google's claims will need to be assessed against independent benchmarks, real-world performance and the costs businesses incur when deploying the model at scale.

For companies considering AI adoption, the key question is no longer only which model can answer a prompt best. It is which model can reliably complete useful work, reduce costs and deliver results that justify its price.