What happened
On Oct 13, 2026, Google made Gemini 4 Argon broadly available. Pro and Ultra subscribers get it in the Gemini app, and Ultra subscribers get it in Google Antigravity.
Developers and enterprises can reach it through the API in AI Studio and the Gemini platform, which Google describes as global availability.
Ultra was the only app tier with access when Argon first appeared in September. Pro subscribers were waiting at the door. Now they are in.
Google pitches Argon at heavy-duty work: coding, finance, legal, cybersecurity and science. It also promises a change in manners. Per Google, the model answers more directly with less hedging, avoids repeating itself, and follows style instructions more reliably than any earlier Gemini.
Google says cached input tokens come at 95% off. The price is labelled introductory, so read the footnote before you build a business on it.
On performance, Google says Argon ranks #1 on the Vals Index, a benchmark covering finance, coding, legal and tax work. That is Google grading its own launch, so treat it as a claim until independent tests pile up.
The app also gets new tricks. Google says Argon shares progress more often on longer tasks. Gemini on macOS lets you switch between chatting and coding, with direct access to Antigravity's agentic coding interface for building apps and web experiences.
Why it matters
A frontier model behind a premium paywall is a demo. A frontier model on the cheaper Pro tier and in the API is a product. That is the shift here.
The pitch is also about temperament, not just raw scores. Less hedging and fewer clichés sounds minor until you remember how many chatbot answers read like a cautious intern covering every base. If Argon really cuts that, daily use gets less irritating.
Then there is the money. Google is putting a low introductory token price on a model it says tops a cross-industry benchmark. Rivals will notice, and developers comparing bills will notice faster.
The robots are not just getting smarter, sapiens. They are getting more direct about it.
What to do with it
- Check your tier.If you pay for AI Pro, open the Gemini app and see whether Argon is on offer.
- Run your own test.Feed it a real task from your job, such as a contract, a spreadsheet or a bug, and compare it with your current model.
- Price a pilot.If you build with APIs, estimate costs at $2 and $10 per million tokens, then check what happens when the introductory price ends.
- Treat the ranking as marketing.The Vals Index claim is Google's, so wait for independent results before switching anything critical.
Keep up, sapiens. The model already stopped hedging on your behalf.


