Google has introduced Gemini 4 Argon, a new AI model designed to handle longer, more complex tasks with far less human guidance. Its standout feature is a massive 1-million-token output window, a major jump from the previous 64,000-token limit. This expanded capacity allows the model to generate much longer responses, maintain context over extended workflows, and complete complicated projects without requiring users to split instructions into multiple smaller prompts.
The larger output window is especially important for advanced AI agents. Instead of responding to one short request at a time, Gemini 4 Argon is built to plan, reason, analyze information, execute steps, test results, and refine its work over a longer period. In practical terms, this means the model could take on large coding projects, deep research tasks, technical analysis, or multi-stage problem-solving with greater independence.
Google is positioning Gemini 4 Argon as another step toward more capable autonomous AI systems. An AI agent powered by this model could receive a broad task, create a detailed plan, inspect code or data, perform intermediate actions, identify mistakes, make corrections, and continue working until the objective is complete. This could reduce the need for constant user intervention and make AI tools more useful for developers, researchers, and enterprise teams.
The company says Gemini 4 Argon is already being used internally across several demanding technical areas. According to Google, Argon agents are helping quantum computing researchers optimize spacetime resources, improving memory efficiency in data centers, converting C and C++ code to Rust, and making Rust-based video decoders faster. These examples suggest that the model is aimed not only at general productivity but also at highly technical workloads where long-context reasoning and iterative problem-solving are essential.
Gemini 4 Argon is not publicly available at launch. Google plans to begin access with trusted testers and cybersecurity specialists before expanding availability to developers, businesses, and consumers. Early access is expected to come first for paying API customers and Google AI Ultra subscribers.
Pricing has also been outlined. During the introductory period, Gemini 4 Argon will cost $2 per 1 million input tokens and $10 per 1 million output tokens. After that period ends, the price is expected to increase to $4 per 1 million input tokens and $20 per 1 million output tokens. Cached input tokens will be offered at a 95% discount, which could help reduce costs for users running repeated or large-scale workloads.
With its 1-million-token output capability, Gemini 4 Argon could become one of Google’s most important AI models for long-running autonomous tasks. If it performs as described, it may significantly change how developers, researchers, and businesses use AI for complex projects that require planning, coding, analysis, and continuous refinement.
