Google has launched Gemini 4 Argon, its latest frontier artificial intelligence model, and calls it its most powerful model yet. The company says Argon improves performance across coding, professional work, cybersecurity, and complex workflows. Google CEO Sundar Pichai introduced the model while addressing growing interest in the company’s next AI generation. He said Argon delivers advanced performance in software engineering, cyber defence, and complex tasks. Google teams have already started using the model for coding and quantum computing projects. The launch also marks the first time Google has used the Argon name for one of its AI models. Google designed the system to handle demanding tasks that require deeper reasoning and longer responses. The company has initially limited access while it continues testing the model and its safety measures.
Also Read : Swara Bhasker Criticises Padmaavat’s Jauhar Sequence, Questions Its Message About Women
Google Next-Generation AI Model
Gemini 4 Argon features a one-million-token output limit, which Google says supports longer and more complex workflows. The model can generate hundreds of thousands of tokens during extended reasoning tasks. Google has also published benchmark results showing Argon competing with other advanced AI models. On the Vals knowledge-work benchmark, Argon achieved a reported score of 68.9 percent. Google reported lower scores of 63.1 percent for GPT-6 Astra and 65.8 percent for Claude Fable 5.1. Argon also achieved 91.9 percent on the Vibe Code Bench coding evaluation. Claude Fable 5.1 scored 90.3 percent, while GPT-6 Astra recorded 89.6 percent. However, Argon did not lead every evaluation, and Google acknowledged that other models performed better on some tests. These results represent Google’s own published comparisons and may not capture every real-world use case.
Cybersecurity forms another major part of Gemini 4 Argon’s capabilities. Google says the model can identify, validate, and patch critical software vulnerabilities. The company has already tested Argon through its cybersecurity operations with Wiz. During an early demonstration, Argon reportedly discovered a critical vulnerability in healthcare software. The vulnerability could expose sensitive personal information across healthcare systems used by hospitals. Google said earlier frontier AI models had failed to identify the same issue. The company sees these capabilities as useful for security researchers and cyber defence teams. However, powerful cybersecurity tools require careful controls because attackers could also misuse advanced AI systems. Google has therefore introduced additional safeguards during Argon’s initial deployment. The company plans to use feedback from trusted cybersecurity professionals before expanding access further.
Also Read : From Farms to Factories: How AI Is Transforming India’s Food Supply Chain
Advanced Cybersecurity and Limited Rollout
Google has started Gemini 4 Argon with a limited release for selected users and organisations. The company currently provides access to trusted cyber defenders through its Fairwind programme and the US government. Google says this controlled approach will help teams evaluate the model’s capabilities and safety. The company plans to expand access to paid API customers and Google AI Ultra subscribers. Google will use feedback from early testers to improve Argon before wider availability. The company says it wants to expand access as quickly as possible while maintaining appropriate safeguards. This staged rollout also allows Google to monitor how developers and security researchers use the model. The approach reflects the growing focus on responsible deployment for increasingly capable AI systems. Google has not yet opened Argon to all consumers through its regular AI products.
Google will initially offer Gemini 4 Argon at an introductory API price for developers and businesses. The company plans to charge $2 per million input tokens and $10 per million output tokens. Google will also provide a major discount on cached input tokens during the introductory period. After the introductory pricing ends, the company plans to charge $4 per million input tokens. The output price will also increase to $20 per million tokens after that period. Google expects developers to use Argon for software engineering, professional workflows, cybersecurity, and other complex applications. The model’s long-context capabilities could also support tasks requiring extensive reasoning and large amounts of generated content. Its limited launch gives Google time to evaluate performance before opening the system more widely. As testing continues, developers and researchers will provide feedback that could shape Argon’s future availability and capabilities.
Also Read : Swara Bhasker Criticises Padmaavat’s Jauhar Sequence, Questions Its Message About Women