Google Gemini 4 Argon offers 1M token context with upcoming price double

The Argon rollout and specs

Google has officially released Gemini 4 Argon, making it available worldwide through the Gemini API. The model is identified by the string gemini-4-argon for developers who want to integrate it into their own applications. It is also accessible to consumers who subscribe to Google AI Ultra or Pro plans.

The primary technical highlight of Argon is its unified context window of 1,048,576 tokens. This means that the total amount of information the model can process at once—including both the user's prompt and the model's response—is capped at this specific number. This allows for the inclusion of large research collections or multi-hour transcripts in a single request.

While the total limit is high, the default synchronous API output is capped at 65,536 tokens to manage latency. Users who require longer generations, such as book-length drafts or large code migrations, can opt for a long-decode mode. This mode allows the model to reach the full 1,048,576-token output ceiling if necessary.

Pricing tiers and the December cliff

The model currently sits in an introductory pricing period that ends on November 30, 2026. During this window, the cost is $2 per 1 million input tokens and $10 per 1 million output tokens. Google also offers a 95% discount on cached input, which brings that specific cost down to $0.10 per 1 million tokens.

Starting December 1, 2026, every price point for the model will double. The input cost will rise to $4 per 1 million tokens, and the output cost will jump to $20 per 1 million tokens. The cached input discount will remain at 95%, meaning the price for cached tokens will move to $0.20 per 1 million tokens.

This uniform doubling means that the ratios between different types of costs will stay the same. For example, output tokens will continue to cost five times as much as fresh input tokens. This transparency allows budget owners to plan for the steady-state costs ahead of time, provided they account for the 2x increase in their forecasts.

Thinking levels and cost management

Gemini 4 Argon introduces three tunable thinking levels: low, medium, and high. These settings allow developers to control how much internal reasoning the model performs before it provides a final answer. Higher settings generally provide more depth for complex problems but require more computational effort from the model.

A critical detail for users is that thinking tokens are billed at the same rate as output tokens. This means that choosing a high thinking level can significantly inflate the cost of a request, even if the final answer provided to the user is relatively short. Developers must balance the need for deep reasoning against the potential for high costs on every call.

To manage these costs, Google suggests starting with the low thinking level for simple tasks like classification or routing. Medium can be reserved for general drafting, while high should be saved for multi-step problems where a mistake is costly. Teams should log their token counts at each level to understand how specific tasks impact their monthly bills.

Who can access the model

Access to Gemini 4 Argon is segmented into three primary paths depending on the user's needs. Developers and enterprises can access it directly through the Gemini API using a billing account. This is the primary route for building custom tools, such as multilingual contract review systems or government form processors.

For individual consumers and prosumers, the model is unlocked through existing Google AI Ultra or Pro subscriptions. This means it is integrated into the consumer-facing Gemini experiences rather than being a standalone API product. This distinction is important for users trying to determine if they can use the model on a free account.

There is also a narrow path for eligible scientists and researchers who can apply for a limited six-month free window. However, this is not an open tier for the general public. For those in regions like Southeast Asia, Google confirms global availability but does not specify local currency billing or specific regional data residency terms.

Strategic implications for builders

The combination of a million-token window and the current introductory pricing makes it easier for small teams to build document-heavy products. Because the large context window reduces the need for complex retrieval pipelines, smaller developers can build more sophisticated tools without needing heavy infrastructure. This is particularly useful for processing large archives or long-form content.

However, the upcoming price doubling in December creates a specific risk for startups. Those who price their products based on the current $2 and $10 rates may see their profit margins squeezed when the costs jump to $4 and $20. This timing risk is a major factor for small builders who may need to adjust their subscription prices before the end of the year.

Google's strategy of using introductory pricing followed by a step-up is a common industry move to gather usage data and size demand. By publishing the exact dates and numbers for the price increase, Google allows buyers to plan their migrations. The ultimate success of Argon will depend on whether its recall quality and reasoning depth justify the premium price point compared to rivals.