What changed
Google announced Gemini 4 Argon for long-running reasoning, software engineering, and professional workflows. Its September 30 post says the initial rollout is limited to a set of trusted cyber defenders in the Fairwind Program, with wider access dependent on further testing and safeguards.
The announcement specifies an output limit of one million tokens. That figure describes output capacity, not the input context window. Google also presents internal examples and benchmark results; those are the provider's evaluations, not SaveMyToken tests. The announcement does not establish general consumer or API availability.
What it means for your workflow
SaveMyToken editorial take
Keep availability separate from capability when shortlisting models. An announced model is only a practical option once your account can access the version and tools your task needs.