Google has begun rolling out Gemini 4 Argon, its newest frontier model, to paying users of the Gemini app and to developers with paid API access. In an announcement dated October 13, the company said AI Pro and AI Ultra subscribers could start using Argon that day. Google first introduced the model at the end of September and gave early access only to a small group of cybersecurity partners. Broad availability puts its top model in front of a much larger audience as the company tries to close a gap with OpenAI and Anthropic.
Who gets access, and where.
Access depends on the product. Google's announcement says the Gemini app and AI Mode are open to AI Pro and AI Ultra subscribers. The Antigravity desktop app and command-line tool, along with AI Studio and the Canary channel of Android Studio, are limited to AI Ultra, which 9to5Google lists at $99.99 and above. Google describes the rollout as starting, so access may vary from one account to another while it spreads across products. The announcement did not say which countries are covered.
Developers and pricing.
Developers with paid Gemini API access can call Argon now. Google is charging introductory rates of $2 per million input tokens and $10 per million output tokens. Those rates end on November 30, after which the price rises to $4 per million input tokens and $20 per million output tokens, according to 9to5Google. Teams that build on the model should budget for that increase before the end of the year, especially if their workloads generate long outputs.
What Google has claimed.
Google calls Argon its "most performant model yet built for complex workloads." Reuters, reporting on the September announcement, said Google's own results show Argon ahead of OpenAI's Astra and Anthropic's Opus on several benchmarks, while trailing on two of four coding tests. The October 13 post describes "major improvements in coding" but does not name benchmarks or publish scores, so the detailed comparisons come from the earlier coverage rather than from the new release.
The launch follows a difficult stretch for Google's AI programme. Reuters reported that Google no longer plans to release Gemini 3.5 Pro, which Sundar Pichai had said in June would arrive. The company also overhauled its DeepMind lab, and several Gemini model leaders left. Argon is Google's attempt to show the model line is back on track after those delays.
What is still unknown.
Several points remain open. The announcement did not include a full benchmark table, and it did not say how long the staggered rollout will take. It is also unclear whether usage limits for coding tools will tighten once the introductory API pricing ends. Independent testers will need time to check the performance claims before anyone can say how Argon compares with the latest models from OpenAI and Anthropic on real work.
For a team deciding whether to adopt the model, the practical step is to test Argon on its own prompts before the introductory price ends. Distributed teams can coordinate those tests across regions with the free time zone converter at /en/time-zone-converter, which helps avoid scheduling a review during someone's night.