Google on October 13 widened access to Gemini 4 Argon, its newest frontier model, moving it beyond the vetted cyber defenders who had exclusive access since the September 30 announcement. According to Google's own availability post, Argon is now available in the Gemini app for Google AI Pro and AI Ultra subscribers.

Where it is available

Google says the model is also reachable through its business and developer channels, including the Gemini API platform, Google AI Studio and Google Antigravity. Secondary coverage reports that AI Studio and the Antigravity desktop app and CLI require an AI Ultra subscription, and that the API identifier is gemini-4-argon on the paid tier. We could not independently confirm those details against Google's developer documentation, so developers should check the current docs before building on them.

Google describes the rollout as gradual and global, so the model may take some hours to appear in individual accounts.

The same post says Antigravity's agentic coding interface is coming to the Gemini macOS app, and that a chat-to-tasks toggle has been added. Antigravity is the agent-first development environment Google has been building out around its Gemini models.

Pricing

Google has listed introductory API pricing of $2 per million input tokens and $10 per million output tokens through November 30, 2026, with standard prices of $4 and $20 per million tokens afterward and a 95% discount on cached input tokens. Those figures match what Google published at the September 30 announcement; the change this week is that more developers can actually use them.

What to watch

At launch Google reported leading or tying on most of the benchmarks it disclosed, including 77.9% on DeepSWE v1.1, while OpenAI's GPT-6 Astra still led on some tests. Those are vendor-reported results. With broader access, independent evaluations of Argon on real coding and agentic workloads should now be possible, and they will show whether the claimed lead holds outside Google's chosen benchmarks. The introductory price window also gives teams until the end of November to test the model at half the standard rate.