Gemini 4 Argon is reaching paying users first, through the Gemini app and Antigravity. Developers using the API get half-price rates through November, according to Ars Technica. Free access has not been confirmed.
What is Gemini 4 Argon?
Gemini 4 Argon is a new model in Google’s Gemini family of large language models. Google uses the Gemini name for its consumer chatbot app and for the models behind it, which developers can also use through an application programming interface (API). Ars Technica reports that Argon has started rolling out. Little else has been confirmed. Its benchmark results, context window, supported input types and position against other Gemini 4 variants are not known. “Argon” looks like a variant name inside the Gemini 4 generation, not a separate product line, but Google’s naming scheme for this generation has not been fully explained. Until official documentation describes what Argon is meant for, it is best treated as a new option to test, not an automatic upgrade.
Where can the model be used?
The report names two places: the Gemini app and Antigravity. The Gemini app is Google’s general-purpose assistant, available on the web and on mobile devices, where people type or speak requests. Antigravity is Google’s software development environment, built around AI agents that can plan and carry out coding tasks inside a project. The model is also available to developers through the API, which is how the discounted rates apply. It is not known whether Argon will reach other Google products, such as productivity apps or search features, or when that might happen. Because the release is described as a rollout, some eligible users may not see the model straight away.
Why is access limited to paying users?
Ars Technica reports that the rollout only covers people who pay. Google has not given a reason in the material available here, but limiting new models to subscribers is common across the industry. Running large models is expensive. A paid tier lets a provider control demand during the early weeks and gives subscribers a concrete reason to keep paying. A gradual release also lets the company watch reliability and usage before opening access more widely. Whether Argon will later reach free users, and on what terms, has not been announced. Anyone relying on the free version of the Gemini app should assume they cannot use Argon for now.
Which subscription do you need?
This is the main detail missing from the available reporting. Google sells more than one paid plan for its AI services. The source material does not say whether Argon comes with every paid plan or only the higher-priced ones. It also does not say whether usage caps differ between plans. Before subscribing or upgrading, check Google’s own plan comparison page and look for Argon by name in the list of included models. If the page does not mention it, ask support or wait for confirmation. Paying for a plan on the assumption that it includes a particular model can be a costly mistake, especially if the plan renews automatically.
How do you check whether Argon is available on your account?
Because the release is staged, having the right plan may not be enough on the first day. In the Gemini app, open the model selector, usually a drop-down menu near the input box, and see whether Argon appears in the list. In Antigravity, look in the settings or the agent configuration for the available models. For API users, the model list endpoint or the developer console shows which model identifiers your project can call. If Argon is missing, sign out and back in, or update the app. If it is still missing, wait a few days before contacting support, because rollouts often reach accounts in waves.
What is the API discount, and how long does it last?
Ars Technica reports that developers will pay 50 per cent less than the standard API rates for Argon through November. API use is normally billed per token, a unit of text roughly equal to part of a word, with separate prices for input and output. A 50 per cent discount halves those charges for the period it applies. The reporting does not give the full-price rates. It is also not clear whether the discount covers every type of request, such as batch jobs or cached input. Check the official pricing page for the exact per-token figures, and note the point at which billing returns to the standard rate.
How should developers use the discount period?
The discount makes this a cheaper time to evaluate the model. One practical approach is to put together a fixed set of real tasks from your own application and run them through Argon and through the model you currently use. Then compare output quality, response time and token usage. Record what each run costs at the discounted rate, then double the figure to estimate what it will cost after the discount ends. That second number is the one to plan around. Set spending limits or budget alerts in your cloud billing settings before running large tests, because cheaper rates can still add up quickly at volume.
Is it worth switching from an older Gemini model?
That depends on your workload, and there is not enough public evidence to give a general answer. Newer models are not always better at every task, and they can behave differently from the model a prompt was written for. For developers, switching can change output formats, refusal behaviour and costs, and any of those can break an existing integration. A cautious approach is to keep your current model in production while you test Argon on the side, then move traffic over gradually if the results hold up. For individual users of the Gemini app, try Argon on the tasks you do most often and see whether you can tell a difference.
What should Antigravity users know before relying on it?
Agent-based coding tools can make many model calls for a single request, because the agent plans, edits files, runs commands and checks results. Any limits on Argon use therefore matter more in Antigravity than in a chat window. The reporting does not describe usage limits for Argon in Antigravity. Before running a long agent session, check whether your plan has a request quota and what happens when you reach it. Review changes made by any agent before committing them, regardless of the model, and use version control so an unwanted edit can be undone easily.
What is still unknown about the release?
Several practical details are missing from the information available. There is no confirmed list of which paid plans include Argon. No timetable for free-tier access has been given. The standard API prices that will apply after November have not been stated here, and there is no independent assessment of the model’s performance. It is also unclear how Argon relates to any other Gemini 4 variants. Readers deciding whether to pay should rely on Google’s official documentation and pricing pages, which should be updated as the rollout continues. Independent testing by developers and reviewers will probably give a clearer picture of the model’s strengths in the weeks after release.
Sources and further reading
- Ars Technica: report on the paid-only Argon rollout and the temporary API discount
- Google official product pages: plan comparisons for the Gemini app subscriptions
- Google developer documentation: model lists, API pricing and rate limits
- Antigravity product documentation: model selection and agent configuration
Surfaced from the rss:arstechnica signal “AI model paid rollout”. AI-assisted draft, editorially reviewed.
