Google unveiled Gemini 4 Argon in late September. On paper, this model seemed Google’s most powerful response to the best models from OpenAI and Anthropic, but at launch almost no one could use it. At that time Argon was only available to a small group of testers, trusted partners, and government projects, and Google continued testing the model before a wider release. Now, two weeks later, that restriction has been lifted.
Who has access to Argon?
Gemini 4 Argon is now available globally via the paid tier of the Gemini API, AI Studio, and Google’s Gemini platform for developers and companies. Google AI Ultra subscribers can also access it via Antigravity, and Pro and Ultra subscribers can use Argon in the Gemini app and the AI search mode.
This was not new; Google had previously announced that paid API customers and Google AI Ultra subscribers would be the first groups to gain access to the model.
What is Argon built for?
According to Google, Argon is designed for heavier-duty tasks. The model is offered for coding, finance, legal work, cybersecurity, science, and other domains where the model should focus on a complex task for a while instead of providing a quick answer.
Google says Argon ranks first on the Vals index, which measures AI performance in areas such as finance, coding, legal, and tax.
Competitive pricing
API pricing is competitive as well. Google has priced Argon at an initial $2 for each one million input tokens and $10 for each one million output tokens. Cached input tokens also receive a 95 percent discount.
Use cases
Compared to the launch, Google now presents more real-world examples. According to the company, Argon helped increase agentic code submissions within Google by over 35 percent and also reduced rollbacks to previous versions by about a third. Chrome engineers used the model in a large refactor covering hundreds of files in the nearly 40-million-line Chromium codebase.
Since then, Argon has been used to optimize Google’s agent development kit, speed up the xTB quantum chemistry package, improve Gemma 4 inference, and assist researchers in cybersecurity and scientific fields.
Free users still left out
Despite the expanded access, free Gemini users still cannot use this model. For regular users, Argon remains an exclusive advantage for Google AI Pro or Ultra subscribers for now.
Additional notes
Token concept: AI API usage costs are typically calculated based on the number of input and output tokens. Tokens are small units of text and are roughly equivalent to parts of a word.
Cache usefulness: The 95 percent discount on cached tokens can noticeably reduce costs for applications where a large portion of requests are repetitive (such as organizational assistants with fixed prompts).
Recommendation to users: Before upgrading to a paid subscription, review the terms and limits of each plan on Google’s official page.