On October 13, 2026, Google officially expanded global access to its frontier model, Gemini 4 Argon, for developers and individual users. This move comes nearly two weeks after the model was announced on September 30, 2026, when Google restricted testing access exclusively to a group of cybersecurity partners under the Fairwind program.
According to an announcement from Google, individual users subscribed to Google AI Pro or Ultra plans can now experience Argon directly in the Gemini app and via the AI Mode feature in Search. For the developer community and enterprises, the model is available through the paid API on Google AI Studio, Android Studio, and the Google Antigravity agent development platform, which is exclusive to Ultra subscribers.
Introductory Pricing and Cost Structure via API
Google has set an introductory API price of $2 per million input tokens and $10 per million output tokens, accompanied by a 95% discount for cached tokens. This promotional pricing will remain in effect through November 30, 2026, before transitioning to the standard rate of $4 for inputs and $20 for outputs.
The most notable technical advancement in this widespread release is a unified output capacity ceiling of up to 1 million tokens. The company explained that expanding the output limit enables the model to maintain seamless reasoning chains across long-horizon tasks without interruptions or the need to fragment work into intermediate steps.
Real-World Applications and Internal Benchmarks
Google stated that internal engineering teams utilized Argon to execute large-scale codebase refactoring, including decoupling web-page rendering code within the Chromium project, which spans over 40 million lines of code. In the Agent Development Kit for Python, the model automatically suggested optimizations that cut in-memory session preparation time by 40% and accelerated initial library loading speeds by 21%.
In standardized evaluations, Google reported that Argon achieved 77.9% on DeepSWE v1.1 and secured the top spot on the Vals Index, a benchmark assessing task execution performance across economic domains such as finance, law, and software engineering. However, these figures currently rely entirely on self-reported data from Google and will require further independent verification by the community in real-world production environments.
Google’s accelerated timeline for bringing Argon into commercial channels indicates that the company is actively striving to translate frontier research capabilities into mainstream products, aiming to apply direct competitive pressure to rivals such as GPT-6.1 Sol and Claude in the deep-reasoning segment. The next key aspect to monitor is whether the actual operational costs at the 1-million-token output threshold will prove economical enough for enterprises to integrate Argon into their long-term operational workflows.