Google has started the general release of Gemini 4 Argon, its most advanced model, a few weeks after announcing it. Developers can use it through the API, and Pro and Ultra subscribers get it in the app, with an introductory half price until late November. Google highlights complex coding and autonomous tasks, with a limit of up to one million tokens. A new safeguard halts runs when it detects a suspicious workflow.
What it means: Making Gemini 4 Argon generally available only in paid plans and through the API shows Google treating its strongest models as a high-value commercial product rather than a free way to attract users, which fits their high running costs. The focus on complex coding, autonomous tasks and a limit of up to one million tokens means the model is aimed at teams that want to hand it entire projects, large document sets or big code repositories, not just short questions. The mechanism that halts a run when it detects a suspicious workflow is an important detail, because it shows vendors taking on some responsibility for what agents do, but also means a legitimate job may occasionally stop without warning, which has to be planned for in system design. For businesses evaluating the model, the half-price introductory period until late November is a good window for testing on real scenarios, measuring cost per task before any long-term commitment.
Read the original on Ars Technica
Related service: Custom software