Google's October 13 release is available to paying Gemini users and developers through AI Studio. The introductory API prices are $2 per million input tokens and $10 per million output tokens, doubling after November 30. Google also links Argon to task-running features and the Antigravity coding environment on Mac.
An especially relevant test is a bounded task from the AI-spend monitor, Sonos controller or an ESP32 display repo. Have Argon diagnose a small bug and produce tests, then repeat with the same context using another model. Measure whether it actually produces a correct, reviewable patch rather than relying on benchmark headlines. Google also offers an app task mode and Antigravity coding integration on Mac.
Action today: log prompt, code quality, test passes, corrections and actual cost for each model. Model the doubled December pricing before moving any production API work.
Recommendation: run the same small repo task with Argon and a current model, recording test results, human corrections, token usage and cost. The December rate matters more than a one-day benchmark victory.