Google’s newest flagship model is moving beyond its most expensive subscription. Gemini 4 Argon, announced at the end of September and initially restricted to AI Ultra, is now rolling out in the Gemini app to AI Pro subscribers as well, while developer access through Google’s coding tools remains firmly locked behind the top tier.
Inside the app, Google is pitching Argon on reasoning and its handling of complex, multi-step work. Much of the messaging, though, focuses on personality fixes. The company says the model gives more direct answers, hedges less, avoids repeating itself and cuts out clichéd phrasing, while following style instructions more consistently than earlier Gemini releases. For longer jobs, it is supposed to report progress more frequently. Those claims target some of the most common frustrations with chatbots generally, but they’re also the kind of improvements that only hold up through sustained everyday use.
The app itself is changing to match. Mac users can now jump from chatting to coding inside the Gemini client, with direct access to Antigravity’s agentic coding interface for building apps and web projects. A new toggle in the main chat window switches between chat and tasks, the feature previously called Spark. The model picker for paying subscribers now lists 3.5 Flash-Lite, 3.8 Flash, 3.1 Pro and 4 Argon, a lineup that’s getting harder for casual users to make sense of. Argon is also available in AI Mode for Pro and Ultra subscribers.
For developers, the picture is more restrictive. Using Argon in the Antigravity desktop app, Antigravity CLI, AI Studio and the Canary build of Android Studio requires AI Ultra. Google is backing that up with internal success stories, saying Argon helped the Chrome team untangle a legacy portion of Chromium in weeks rather than months, nearly tripled inference throughput for a Gemma 4 model, and halved CPU usage on YouTube’s comments frontend. These are impressive figures, but they’re self-reported results from Google engineers working on Google’s own codebases, not independent benchmarks.
API pricing is where planning matters. Argon starts at $2 per million input tokens and $10 per million output tokens, but that’s an introductory rate ending November 30. After that, prices double to $4 and $20 respectively, a sharp increase for teams building workflows around the model during the launch window.
Google is also offering 10,000 scientists six months of free access to Antigravity’s highest usage tier, with applications now open. It’s a smart way to seed research use cases, and the real test will be whether Argon’s productivity claims survive contact with codebases Google doesn’t own.
