Google presents Gemini 4 Argon, the new artificial intelligence model

Among the new features is a limit of one million tokens on exit, among the highest in the sector. The launch comes just days after new models unveiled by OpenAI and Anthropic, reigniting competition in artificial intelligence

Google also enters the new phase of the artificial intelligence competition with Gemini 4 Argon, an advanced model that will not be available to the public for the moment. The company decided to initially distribute it to a select group of cybersecurity experts and use it internally, adopting a gradual release to test its reliability and security.

Performance and processing capacity

According to Google, Gemini 4 Argon is designed to handle particularly complex tasks, from software development to professional applications in the legal, financial and cybersecurity sectors. Among the main innovations is the increase in the limit of outgoing tokens up to one million, one of the highest values ​​in the sector. Tokens represent units of text that AI models use to process and generate content.

A gradual release for security reasons

Google explained that the launch of frontier technologies requires a gradual path. The company is participating in the US government's voluntary program that allows access to models before official release, while gathering feedback from early tests to strengthen security measures. The decision comes after a series of episodes and reports related to the security of AI systems recorded in recent months and which involved various players in the sector, including Google itself.