Gemini 4 Argon, the AI model Google unveiled at the end of September and billed as the most capable system it has ever built, is no longer locked behind enterprise doors. When it launched, the model was reserved for corporate customers and software developers only.

That changed with an announcement made yesterday evening: Google confirmed that Argon has now entered general availability.

Subscribers get first access

gemi1

According to the company, anyone on an AI Pro or AI Ultra plan can start using the model immediately. Google also hinted that its user base could expand further before the year is out. On the developer side, access runs through the paid API tiers of Google AI Studio, Android Studio and Google Antigravity.

Built for demanding work

Argon stands out on tasks that demand deep expertise and many steps, such as finance, law, cybersecurity and science. The model took the top spot in the Vals Index, an evaluation focused on economics and business, and it has become a daily tool for Google's own engineering teams on code development and architecture modernization work. Since Argon was deployed in the Antigravity environment, agent-based code submissions rose by more than 35%, while the amount of reverted code dropped by roughly a third, the company shared.

Safety hasn't been overlooked

Alongside its strong reasoning capabilities, Google points to the model's protective guardrails. Argon has been trained to refuse requests to produce malware or instructions involving dangerous biological materials. When it hits a blocking point, the model turns to human approval rather than making risky guesses. In most tests run against rivals GPT-6 Astra and Claude Fable 5.1, Argon came out ahead.