What changed on October 13
Google’s Gemini API release notes now list Gemini 4 Argon as generally available under the model name gemini-4-argon. Its developer guide identifies the Gemini API and Google AI Studio as access routes. This updates our earlier coverage of the September 30 announcement, which described an initial restricted rollout to trusted cyber defenders.
For the Gemini app, Google’s October 13 release notes announce a global rollout to Google AI Pro and Ultra subscribers. A rollout can take time; the announcement does not mean every account already shows the model.
For someone learning AI
Google describes Argon as a reasoning model for difficult coding, agent, mathematics, and science tasks. The model page lists text, image, video, audio, and PDF inputs, with text output. That does not mean it generates images or voices: those output capabilities and the Live API are listed as unsupported.
Our takeaway: a more capable model can still make mistakes. Start with a task whose answer you can check, and keep the difference between the model’s abilities and an app’s features in mind. We have not run a hands-on Argon test.
A detail for the Workbench
Google’s documentation describes one shared window for input, thinking, and output, totaling 1,048,576 tokens. The default synchronous output limit is 65,536 tokens. Longer generation requires the documented long-decode workflow; the million-token headline is not an automatic million-token answer allowance on top of your input.
In the API, thinking levels include low, medium, and high, with medium the default. Developers should check current usage charges and output limits before enabling long responses. This website has not enabled an Argon endpoint or connected visitor questions to a paid API.
How this briefing changed
We updated our earlier restricted-access headline after checking the October 13 developer and app release notes. One inconsistency remains: the developer guide still includes a Google AI Pro exclusion, while the app release notes and Apps Help list Pro and Ultra access. We cite each product’s documentation rather than promising account-level eligibility. Check your own model picker and current plan details.
The September announcement remains linked for historical context. The outside evaluation concerns that earlier release; it does not establish current app access. Our own recorded Gemini example uses Flash-Lite, not Argon.