{"id":14421,"date":"2026-10-03T06:00:44","date_gmt":"2026-10-03T06:00:44","guid":{"rendered":"https:\/\/krikis.lt\/google-gemini-4-argon-naujas-modelis\/"},"modified":"2026-10-03T06:00:44","modified_gmt":"2026-10-03T06:00:44","slug":"google-gemini-4-argon-naujas-modelis","status":"publish","type":"post","link":"https:\/\/krikis.lt\/en\/google-gemini-4-argon-new-model\/","title":{"rendered":"Google launches Gemini 4 Argon, a new AI model for cybersecurity and complex tasks"},"content":{"rendered":"

Google officially announced its new artificial intelligence platform, Gemini 4 Argon, on September 30, 2026. The model, which is first available to trusted cybersecurity professionals through the Fairwind program, is designed to solve extremely complex problems in software engineering, finance, law, and, of course, cybersecurity. Although the model is not yet available to the general public, its technical features and early overview of the scope of application are already generating a lot of discussion in the technology community.<\/p>\n

Long context processing and huge token limit<\/h2>\n
\"Data<\/figure>\n

One of the most significant improvements in Gemini 4 Argon is the expansion of the token limit to 1 million, compared to the 64,000 token limit in previous Gemini generations. This allows the model to maintain context over extremely long work sequences, such as auditing an entire codebase or executing complex business processes that take longer than traditional AI systems can handle. This feature is especially important for those working with large data sets or requiring in-depth analysis without fragmenting.<\/p>\n

Cybersecurity is the main goal of the Fairwind program<\/h2>\n
\"Cybersecurity<\/figure>\n

\u201eThe Fairwind program is Google\u2019s initiative to provide trusted cybersecurity professionals with access to the latest AI tools. According to sources from Il Sole 24 Ore and Ghacks Technology News, Argon has already helped identify a critical vulnerability that could have exposed patient data in a large global hospital software. This vulnerability analysis was conducted in collaboration with security firm Wiz through its Scan for Good initiative. Google says that previous models missed the flaw, and Argon was able to detect it autonomously.<\/p>\n

Software Engineering and Performance Optimization<\/h2>\n

Google also highlights that Argon can significantly reduce resource requirements during complex calculations. For example, the model proposed a solution in a matter of minutes that reduced the resource usage of 40 % critical quantum computing components. Argon also optimized memory management on servers, freeing up more than 300 TB of RAM, and this number is predicted to reach up to 1 PB.<\/p>\n

Pricing and availability<\/h2>\n

While the model is not yet available to the general public, Google has already announced preliminary pricing: $2 per million input tokens and $10 per million output tokens. This price could change after the introductory period, but there is currently no confirmed information on the final pricing model or final public access date.<\/p>\n

Competitive context: Google vs. OpenAI and Anthropic<\/h2>\n

The Gemini 4 Argon release comes at the same time as Anthropic introduced Claude Opus and Sonnet 5.5, and OpenAI introduced ChatGPT-6. This \u201eheat\u201c between the three big AI players shows that there is an intense battle for leadership in the market, both technologically and commercially. Online-Tech-Tips indicates that Google plans to offer the model to paying API customers and Google AI Ultra subscribers at a later date, but there is no official date yet.<\/p>\n

Security measures and ethics controls<\/h2>\n

Google emphasizes that Argon is restricted from malicious tasks \u2014 the model refuses to execute queries related to cyber, biological, chemical, bacteriological, or nuclear weapons. In addition, the model\u2019s prompt injection index \u2014 0.7 % \u2014 is one of the lowest among competitors, including Anthropic, OpenAI, and even Elon Musk\u2019s Grok model. This shows that Google has invested in security mechanisms that protect against attempts to circumvent the model\u2019s limitations.<\/p>\n

Conclusions<\/h2>\n

Gemini 4 Argon is an ambitious step forward in AI, especially when it comes to long, multi-step tasks and cybersecurity solutions. Although the model is currently only available to a limited number of partners, its technical capabilities \u2013 a huge token limit, autonomous vulnerability detection and efficiency improvements \u2013 could have a significant impact on both Google\u2019s internal operations and the broader AI ecosystem. We\u2019ll see how this model takes hold in the market and what real-world impact it will have on competition with OpenAI and Anthropic.<\/p>\n

\u2014<\/p>\n

CTA:<\/strong> To learn how Gemini 4 Argon could strengthen your organization's cybersecurity, contact Krikis IT, specialists in AI and automation.<\/p>\n

\n

To learn how Gemini 4 Argon could strengthen your organization's cybersecurity, contact Krikis IT, specialists in AI and automation.<\/p>\n<\/div>\n

Sources<\/h2>\n