Google has announced Gemini 4 Argon, a new advanced artificial intelligence model designed to handle complex and lengthy tasks across software development, business and cybersecurity.
According to Google, Gemini 4 Argon is built for what the company describes as “long-horizon” workflows. Instead of focusing only on short questions and answers, the model is designed to work through complicated tasks that may require extended reasoning and multiple steps.
One of the model’s major features is its expanded output capacity. Gemini 4 Argon can generate up to 1 million tokens in a single trajectory, a significant increase from the previous 64,000-token limit. Google says this larger capacity can help the model work on large codebases and complex projects with greater depth.
Google is already using Argon internally for several technical projects. The company says its AI agents have assisted with software optimization, large-scale code migrations and research involving quantum computing. In one example, Google says Argon helped optimize a quantum computing problem and achieved results above a published baseline.
Cybersecurity is another major focus of Gemini 4 Argon. Google says the model has been designed to identify, verify and help fix software vulnerabilities. The company is initially providing access through its Fairwind Program to a selected group of trusted cybersecurity defenders.
Google is taking a gradual approach because of the model’s advanced capabilities. Before wider release, the company says it is strengthening safeguards against misuse, prompt-injection attacks and situations where an AI system could act beyond a user's intended instructions.
Gemini 4 Argon is not yet broadly available to everyone. Google plans to expand access gradually, with paid API customers and Google AI Ultra subscribers among the groups expected to receive access as the rollout progresses.
The model will launch with an introductory API price of $2 per million input tokens and $10 per million output tokens, according to Google. The company says the new model is intended to support developers, professionals and enterprises working on demanding technical and knowledge-based tasks.