Gemini 4 Argon begins rolling out in Gemini app and Antigravity, but only if you pay
Ars Technica Ryan Whitwam ● Covered by 8 sources
Gemini 4 Argon is rolling out now, but only for paying users. Google says the model’s skills haven’t changed — just the price of entry.
Based on reporting by Ars Technica, Ryan Whitwam — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Google said a few weeks ago that Gemini 4 Argon would need a careful rollout because of the risk of misuse. That caution seems to have lasted about as long as it took to line up the paying customers. The model is starting to reach users today, but not everyone gets the same access, and people on lower-tier plans won’t get the full package.
The capabilities are the same as before. Google is still pointing people to the benchmarks it published late last month, where it framed Argon as unusually strong at reasoning and agentic coding. The company also says the model is trained to reject malicious instructions, and that an automated misalignment monitor can stop a run if it spots a workflow going off course.
Google’s pitch is that those safeguards are needed because the model is so capable. And it is leaning hard on internal examples to make that case. Thousands of Googlers have already used Argon, the company says, and one of the cleaner wins came in YouTube’s comment system, where the model spotted redundant emoji rendering. Removing that extra work cut server load by 45 percent.
The model also found old bugs in Project Wing’s proprietary drone physics simulator, including issues that had been sitting in the codebase since 2018. Google says Argon has also pushed its robotics work forward, with a measurable jump in embodied reasoning. So the public rollout is cautious, but the message is not subtle: Google wants Argon seen as powerful enough to justify both the restrictions and the price tag.
My take — AI-written commentary, not fact-checked reporting
This is the usual premium-AI script: call it dangerous, then sell the danger with a monthly fee. Google is treating safety like a velvet rope, which is a tidy way to make access feel selective instead of limited. The real story is that the model’s value is being marketed through internal wins first, and that’s exactly how these launches are being normalized now.
Read more about this at: Ars Technica